mirrors/cosmopolitan
1
0
Fork 0
mirror of https://github.com/jart/cosmopolitan.git synced 2025-02-01 12:03:41 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
cosmopolitan/libc/calls/read-nt.c

828 lines
25 KiB
C
Raw Normal View History

Initial import
2020-06-15 14:18:57 +00:00
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:2;tab-width:8;coding:utf-8 -*-│
│vi: set net ft=c ts=2 sts=2 sw=2 fenc=utf-8 :vi│
╞══════════════════════════════════════════════════════════════════════════════╡
│ Copyright 2020 Justine Alexandra Roberts Tunney │
│ │
Change license
2020-12-28 01:18:44 +00:00
│ Permission to use, copy, modify, and/or distribute this software for │
│ any purpose with or without fee is hereby granted, provided that the │
│ above copyright notice and this permission notice appear in all copies. │
Initial import
2020-06-15 14:18:57 +00:00
│ │
Change license
2020-12-28 01:18:44 +00:00
│ THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL │
│ WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED │
│ WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE │
│ AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL │
│ DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR │
│ PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER │
│ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR │
│ PERFORMANCE OF THIS SOFTWARE. │
Initial import
2020-06-15 14:18:57 +00:00
╚─────────────────────────────────────────────────────────────────────────────*/
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/atomic.h"
Initial import
2020-06-15 14:18:57 +00:00
#include "libc/calls/internal.h"
Improve synchronization - Fix bugs in kDos2Errno definition - malloc() should now be thread safe - Fix bug in rollup.com header generator - Fix open(O_APPEND) on the New Technology - Fix select() on the New Technology and test it - Work towards refactoring i/o for thread safety - Socket reads and writes on NT now poll for signals - Work towards i/o completion ports on the New Technology - Make read() and write() intermittently check for signals - Blinkenlights keyboard i/o so much better on NT w/ poll() - You can now poll() files and sockets at the same time on NT - Fix bug in appendr() that manifests with dlmalloc footers off
2022-04-15 06:39:48 +00:00
#include "libc/calls/sig.internal.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/calls/state.internal.h"
Reduce build graph by another eight percent
2022-08-13 20:11:56 +00:00
#include "libc/calls/struct/fd.internal.h"
#include "libc/calls/struct/iovec.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/calls/struct/sigset.internal.h"
Reduce Makefile dependencies by 4%
2022-05-23 22:06:11 +00:00
#include "libc/calls/syscall_support-nt.internal.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/cosmo.h"
Improve Windows Console I/O - Blocking read operations on the Windows Console can now EINTR - Blocking read operations on Windows pipes now EINTR more reliably - setitimer() will no longer be inherited across fork() on Windows - It's now possible to use ECHO when the console is in raw mode - The ECHOCTL flag now works correctly on the Windows Console - The ICRNL flag now works correctly on the Windows Console - pread() and pwrite() will now raise ESPIPE on Windows - Opening /dev/tty on Windows is improved (untested) - Overlapped I/O is now implemented in a better way
2023-08-08 11:00:29 +00:00
#include "libc/errno.h"
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
#include "libc/fmt/itoa.h"
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
#include "libc/intrin/atomic.h"
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
#include "libc/intrin/describeflags.internal.h"
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
#include "libc/intrin/dll.h"
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
#include "libc/intrin/kprintf.h"
Get Fat Emacs working in Windows Console
2023-08-18 11:55:57 +00:00
#include "libc/intrin/nomultics.internal.h"
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
#include "libc/intrin/strace.internal.h"
Make improvements for Actually Portable Emacs - Get SIGWINCH working again on the New Technology - Correctly handle O_NOFOLLOW in open() on Windows - Implement synthetic umask() functionality on Windows - Do a better job managing file execute access on Windows - Fill in `st_uid` and `st_gid` with username hash on Windows - Munge UNICODE control pictures into control codes on Windows - Do a better job ensuring Windows console settings are restored - Introduce KPRINTF_LOG environment variable to log kprintf to a file
2023-08-19 13:41:06 +00:00
#include "libc/intrin/weaken.h"
Incorporate more small improvements
2023-07-23 17:57:18 +00:00
#include "libc/macros.internal.h"
Fix issue with pasting text into console
2023-09-21 20:28:52 +00:00
#include "libc/mem/mem.h"
Make improvements for Actually Portable Emacs - Get SIGWINCH working again on the New Technology - Correctly handle O_NOFOLLOW in open() on Windows - Implement synthetic umask() functionality on Windows - Do a better job managing file execute access on Windows - Fill in `st_uid` and `st_gid` with username hash on Windows - Munge UNICODE control pictures into control codes on Windows - Do a better job ensuring Windows console settings are restored - Introduce KPRINTF_LOG environment variable to log kprintf to a file
2023-08-19 13:41:06 +00:00
#include "libc/nt/console.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/nt/createfile.h"
#include "libc/nt/enum/accessmask.h"
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
#include "libc/nt/enum/consolemodeflags.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/nt/enum/creationdisposition.h"
#include "libc/nt/enum/filesharemode.h"
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
#include "libc/nt/enum/vk.h"
Initial import
2020-06-15 14:18:57 +00:00
#include "libc/nt/errors.h"
#include "libc/nt/runtime.h"
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
#include "libc/nt/struct/inputrecord.h"
Improve synchronization - Fix bugs in kDos2Errno definition - malloc() should now be thread safe - Fix bug in rollup.com header generator - Fix open(O_APPEND) on the New Technology - Fix select() on the New Technology and test it - Work towards refactoring i/o for thread safety - Socket reads and writes on NT now poll for signals - Work towards i/o completion ports on the New Technology - Make read() and write() intermittently check for signals - Blinkenlights keyboard i/o so much better on NT w/ poll() - You can now poll() files and sockets at the same time on NT - Fix bug in appendr() that manifests with dlmalloc footers off
2022-04-15 06:39:48 +00:00
#include "libc/nt/synchronization.h"
Improve Windows Console I/O - Blocking read operations on the Windows Console can now EINTR - Blocking read operations on Windows pipes now EINTR more reliably - setitimer() will no longer be inherited across fork() on Windows - It's now possible to use ECHO when the console is in raw mode - The ECHOCTL flag now works correctly on the Windows Console - The ICRNL flag now works correctly on the Windows Console - pread() and pwrite() will now raise ESPIPE on Windows - Opening /dev/tty on Windows is improved (untested) - Overlapped I/O is now implemented in a better way
2023-08-08 11:00:29 +00:00
#include "libc/str/str.h"
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
#include "libc/str/utf16.h"
Support non-blocking i/o across platforms This change introduces new tests for `O_NONBLOCK` and `SOCK_NONBLOCK` to confirm that non-blocking i/o is now working on all supported platforms, including Windows. For example, you can now say on Windows, MacOS, etc.: socket(AF_INET, SOCK_STREAM | SOCK_NONBLOCK, IPPROTO_TCP); To create a non-blocking IPv4 TCP socket. Or you can enable non-blocking i/o on an existing socket / pipe / etc. file descriptor by calling fcntl fcntl(fd, F_SETFL, fcntl(fd, F_GETFL) | O_NONBLOCK); This functionality is polyfilled on older Linux kernels too, e.g. RHEL5. Now that fcntl() support is much better the FIOCLEX / FIONCLEX polyfills for ioctl() have been removed since they're ugly non-POSIX diameond APIs This change fixes a weakness in kprintf() that was causing Windows trace tools to frequently crash.
2023-07-23 09:56:47 +00:00
#include "libc/sysv/consts/o.h"
Get Fat Emacs working in Windows Console
2023-08-18 11:55:57 +00:00
#include "libc/sysv/consts/sig.h"
Initial import
2020-06-15 14:18:57 +00:00
#include "libc/sysv/errfuns.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/thread/posixthread.internal.h"
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
#include "libc/thread/thread.h"
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
#include "libc/thread/tls.h"
Get Fat Emacs working in Windows Console
2023-08-18 11:55:57 +00:00
#ifdef __x86_64__
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
/**
* @fileoverview Cosmopolitan Standard Input
*
* This file implements pollable terminal i/o for Windows consoles. On
* Windows 10 the "virtual terminal processing" feature works great on
* output but their solution for input processing isn't good enough to
* support running Linux programs like Emacs. This polyfill fixes that
* and it most importantly ensures we can poll() standard input, which
* would otherwise have been impossible. We aren't using threads. What
* we do instead is have termios behaviors e.g. canonical mode editing
* happen on demand as a side effect of read/poll/ioctl activity.
*/
struct VirtualKey {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
int vk;
int normal_str;
int shift_str;
int ctrl_str;
int shift_ctrl_str;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
};
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
#define S(s) W(s "\0\0")
#define W(s) (s[3] << 24 | s[2] << 16 | s[1] << 8 | s[0])
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static const struct VirtualKey kVirtualKey[] = {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
{kNtVkUp, S("A"), S("1;2A"), S("1;5A"), S("1;6A")},
{kNtVkDown, S("B"), S("1;2B"), S("1;5B"), S("1;6B")},
{kNtVkRight, S("C"), S("1;2C"), S("1;5C"), S("1;6C")},
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
{kNtVkLeft, S("D"), S("1;2D"), S("1;5D"), S("1;6D")},
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
{kNtVkInsert, S("2~"), S("2;2~"), S("2;5~"), S("2;6~")},
{kNtVkDelete, S("3~"), S("3;2~"), S("3;5~"), S("3;6~")},
{kNtVkHome, S("H"), S("1;2H"), S("1;5H"), S("1;6H")},
{kNtVkEnd, S("F"), S("1;2F"), S("1;5F"), S("1;6F")},
{kNtVkPrior, S("5~"), S("5;2~"), S("5;5~"), S("5;6~")},
{kNtVkNext, S("6~"), S("6;2~"), S("6;5~"), S("6;6~")},
{kNtVkF1, -S("OP"), S("1;2P"), S("11^"), S("1;6P")},
{kNtVkF2, -S("OQ"), S("1;2Q"), S("12^"), S("1;6Q")},
{kNtVkF3, -S("OR"), S("1;2R"), S("13^"), S("1;6R")},
{kNtVkF4, -S("OS"), S("1;2S"), S("14^"), S("1;6S")},
{kNtVkF5, S("15~"), S("28~"), S("15^"), S("28^")},
{kNtVkF6, S("17~"), S("29~"), S("17^"), S("29^")},
{kNtVkF7, S("18~"), S("31~"), S("18^"), S("31^")},
{kNtVkF8, S("19~"), S("32~"), S("19^"), S("32^")},
{kNtVkF9, S("20~"), S("33~"), S("20^"), S("33^")},
{kNtVkF10, S("21~"), S("34~"), S("21^"), S("34^")},
{kNtVkF11, S("23~"), S("23$"), S("23^"), S("23@")},
{kNtVkF12, S("24~"), S("24$"), S("24^"), S("24@")},
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
{0},
};
// TODO: How can we configure `less` to not need this bloat?
static const struct VirtualKey kDecckm[] = {
{kNtVkUp, -S("OA"), -S("OA"), S("A"), S("A")},
{kNtVkDown, -S("OB"), -S("OB"), S("B"), S("B")},
{kNtVkRight, -S("OC"), -S("OC"), S("C"), S("C")},
{kNtVkLeft, -S("OD"), -S("OD"), S("D"), S("D")},
{kNtVkPrior, S("5~"), S("5;2~"), S("5;5~"), S("5;6~")},
{kNtVkNext, S("6~"), S("6;2~"), S("6;5~"), S("6;6~")},
{0},
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
};
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
#define KEYSTROKE_CONTAINER(e) DLL_CONTAINER(struct Keystroke, elem, e)
struct Keystroke {
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
char buf[23];
unsigned char buflen;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
struct Dll elem;
};
struct Keystrokes {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
atomic_uint once;
int64_t cin, cot;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
struct Dll *list;
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
struct Dll *line;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
struct Dll *free;
bool end_of_file;
Make some quick fixes
2023-10-09 00:56:59 +00:00
bool ohno_decckm;
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
unsigned char pc;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
uint16_t utf16hs;
pthread_mutex_t lock;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
const struct VirtualKey *vkt;
Make some quick fixes
2023-10-09 00:56:59 +00:00
struct Keystroke pool[512];
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
};
static struct Keystrokes __keystroke;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
textwindows void __keystroke_wipe(void) {
bzero(&__keystroke, sizeof(__keystroke));
}
static textwindows void OpenConsole(void) {
__keystroke.vkt = kVirtualKey;
__keystroke.cin = CreateFile(u"CONIN$", kNtGenericRead | kNtGenericWrite,
kNtFileShareRead, 0, kNtOpenExisting, 0, 0);
__keystroke.cot = CreateFile(u"CONOUT$", kNtGenericRead | kNtGenericWrite,
kNtFileShareWrite, 0, kNtOpenExisting, 0, 0);
}
static textwindows void InitConsole(void) {
cosmo_once(&__keystroke.once, OpenConsole);
}
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
static textwindows void LockKeystrokes(void) {
pthread_mutex_lock(&__keystroke.lock);
}
static textwindows void UnlockKeystrokes(void) {
pthread_mutex_unlock(&__keystroke.lock);
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
textwindows int64_t GetConsoleInputHandle(void) {
InitConsole();
return __keystroke.cin;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
textwindows int64_t GetConsoleOutputHandle(void) {
InitConsole();
return __keystroke.cot;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static textwindows bool IsMouseModeCommand(int x) {
return x == 1000 || // SET_VT200_MOUSE
x == 1002 || // SET_BTN_EVENT_MOUSE
x == 1006 || // SET_SGR_EXT_MODE_MOUSE
x == 1015; // SET_URXVT_EXT_MODE_MOUSE
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
}
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
static textwindows int GetVirtualKey(uint16_t vk, bool shift, bool ctrl) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
for (int i = 0; __keystroke.vkt[i].vk; ++i) {
if (__keystroke.vkt[i].vk == vk) {
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
if (shift && ctrl) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
return __keystroke.vkt[i].shift_ctrl_str;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
} else if (shift) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
return __keystroke.vkt[i].shift_str;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
} else if (ctrl) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
return __keystroke.vkt[i].ctrl_str;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
} else {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
return __keystroke.vkt[i].normal_str;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
}
}
}
return 0;
}
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
static textwindows int ProcessKeyEvent(const struct NtInputRecord *r, char *p) {
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
uint32_t c = r->Event.KeyEvent.uChar.UnicodeChar;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
uint16_t vk = r->Event.KeyEvent.wVirtualKeyCode;
uint16_t cks = r->Event.KeyEvent.dwControlKeyState;
// ignore keyup events
if (!r->Event.KeyEvent.bKeyDown && (!c || vk != kNtVkMenu)) {
return 0;
}
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
#if 0
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// this code is useful for troubleshooting why keys don't work
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
kprintf("bKeyDown=%hhhd wVirtualKeyCode=%s wVirtualScanCode=%s "
"UnicodeChar=%#x[%#lc] dwControlKeyState=%s\n",
r->Event.KeyEvent.bKeyDown,
DescribeVirtualKeyCode(r->Event.KeyEvent.wVirtualKeyCode),
DescribeVirtualKeyCode(r->Event.KeyEvent.wVirtualScanCode),
r->Event.KeyEvent.uChar.UnicodeChar,
r->Event.KeyEvent.uChar.UnicodeChar,
DescribeControlKeyState(r->Event.KeyEvent.dwControlKeyState));
#endif
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// turn arrow/function keys into vt100/ansi/xterm byte sequences
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
int n = 0;
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
int v = GetVirtualKey(vk, !!(cks & kNtShiftPressed),
!!(cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed)));
if (v) {
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
p[n++] = 033;
if (cks & (kNtLeftAltPressed | kNtRightAltPressed)) {
p[n++] = 033;
}
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
if (v > 0) {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
p[n++] = '[';
} else {
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
v = -v;
}
do p[n++] = v;
while ((v >>= 8));
return n;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// ^/ (crtl+slash) maps to ^_ (ctrl-hyphen) on linux
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
if (vk == kNtVkOem_2 && (cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed))) {
p[n++] = 037;
return n;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// handle cases where win32 doesn't provide character
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
if (!c) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
if (vk == '2' && (cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed))) {
c = 0; // ctrl-2 → "\000"
} else if (isascii(vk) && isdigit(vk) &&
(cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed))) {
c = 030 + (vk - '0'); // e.g. ctrl-3 → "\033"
} else if (isascii(vk) && isgraph(vk) &&
(cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed))) {
c = vk ^ 0100; // e.g. ctrl-alt-b → "\033\002"
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
} else {
return 0;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
}
Improve Windows keyboard translation to Linux
2023-10-04 17:32:54 +00:00
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// convert utf-16 to utf-32
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
if (IsHighSurrogate(c)) {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
__keystroke.utf16hs = c;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
return 0;
}
if (IsLowSurrogate(c)) {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
c = MergeUtf16(__keystroke.utf16hs, c);
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// enter sends \r with raw terminals
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
// make it a multics newline instead
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
if (c == '\r' && !(__ttyconf.magic & kTtyNoCr2Nl)) {
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
c = '\n';
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// ctrl-space (^@) is literally zero
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
if (c == ' ' && (cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed))) {
c = '\0';
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// make backspace (^?) distinguishable from ctrl-h (^H)
Fix some additional Windows TTY issues
2023-10-04 06:53:59 +00:00
if (c == kNtVkBack && !(cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed))) {
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
c = 0177;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// handle ctrl-\ and ctrl-c
// note we define _POSIX_VDISABLE as zero
// tcsetattr() lets anyone reconfigure these keybindings
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
if (c && !(__ttyconf.magic & kTtyNoIsigs)) {
if (c == __ttyconf.vintr) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
return __sig_enqueue(SIGINT);
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
} else if (c == __ttyconf.vquit) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
return __sig_enqueue(SIGQUIT);
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
}
}
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
// handle ctrl-d which generates end-of-file, unless pending line data
// is present, in which case we flush that without the newline instead
if (c && c == __ttyconf.veof && !(__ttyconf.magic & kTtyUncanon)) {
if (dll_is_empty(__keystroke.line)) {
Add siglongjmp() for aarch64
2023-09-21 17:10:20 +00:00
__keystroke.end_of_file = true;
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
} else {
dll_make_last(&__keystroke.list, __keystroke.line);
__keystroke.line = 0;
Add siglongjmp() for aarch64
2023-09-21 17:10:20 +00:00
}
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
return 0;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
}
// insert esc prefix when alt is held
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// for example "h" becomes "\033h" (alt-h)
// if up arrow is "\033[A" then alt-up is "\033\033[A"
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
if ((cks & (kNtLeftAltPressed | kNtRightAltPressed)) &&
r->Event.KeyEvent.bKeyDown) {
p[n++] = 033;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// finally apply thompson-pike varint encoding
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
uint64_t w = tpenc(c);
do p[n++] = w;
while ((w >>= 8));
return n;
}
// To use the tty mouse events feature:
// - write(1, "\e[?1000;1002;1015;1006h") to enable
// - write(1, "\e[?1000;1002;1015;1006l") to disable
// See o//examples/ttyinfo.com and o//tool/viz/life.com
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
static textwindows int ProcessMouseEvent(const struct NtInputRecord *r,
char *b) {
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
char *p = b;
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
unsigned char e = 0;
uint32_t currentbs = __ttyconf.mousebs;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
uint32_t ev = r->Event.MouseEvent.dwEventFlags;
uint32_t bs = r->Event.MouseEvent.dwButtonState;
ev &= kNtMouseMoved | kNtMouseWheeled;
bs &= kNtFromLeft1stButtonPressed | kNtRightmostButtonPressed;
if (ev & kNtMouseWheeled) {
// scroll wheel (unnatural mode)
Fix some todos
2023-10-10 06:12:32 +00:00
bool isup = ((int)r->Event.MouseEvent.dwButtonState >> 16) > 0;
if (__ttyconf.magic & kTtyXtMouse) {
if (r->Event.MouseEvent.dwControlKeyState &
(kNtLeftCtrlPressed | kNtRightCtrlPressed)) {
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
e = isup ? 80 : 81;
} else {
Fix some todos
2023-10-10 06:12:32 +00:00
e = isup ? 64 : 65;
}
goto OutputXtermMouseEvent;
} else if (!(r->Event.MouseEvent.dwControlKeyState &
(kNtShiftPressed | kNtLeftCtrlPressed | kNtRightCtrlPressed |
kNtLeftAltPressed | kNtRightAltPressed))) {
// we disable mouse highlighting when the tty is put in raw mode
// to mouse wheel events with widely understood vt100 arrow keys
*p++ = 033;
*p++ = !__keystroke.ohno_decckm ? '[' : 'O';
if (isup) {
*p++ = 'A';
} else {
*p++ = 'B';
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
}
}
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
} else if ((bs || currentbs) && (__ttyconf.magic & kTtyXtMouse)) {
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
if (bs && (ev & kNtMouseMoved) && currentbs) {
e |= 32; // dragging
}
if ((bs | currentbs) & kNtRightmostButtonPressed) {
e |= 2; // right
}
OutputXtermMouseEvent:
*p++ = 033;
*p++ = '[';
*p++ = '<';
p = FormatInt32(p, e);
*p++ = ';';
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
p = FormatInt32(p, (r->Event.MouseEvent.dwMousePosition.X + 1) & 0x7fff);
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
*p++ = ';';
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
p = FormatInt32(p, (r->Event.MouseEvent.dwMousePosition.Y + 1) & 0x7fff);
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
if (!bs && currentbs) {
*p++ = 'm'; // up
} else {
*p++ = 'M'; // down
}
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
__ttyconf.mousebs = bs;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
}
return p - b;
}
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
static textwindows int ConvertConsoleInputToAnsi(const struct NtInputRecord *r,
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
char p[hasatleast 23]) {
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
switch (r->EventType) {
case kNtKeyEvent:
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
return ProcessKeyEvent(r, p);
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
case kNtMouseEvent:
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
return ProcessMouseEvent(r, p);
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
case kNtWindowBufferSizeEvent:
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
return __sig_enqueue(SIGWINCH);
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
default:
return 0;
}
}
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
static textwindows struct Keystroke *NewKeystroke(void) {
struct Dll *e;
struct Keystroke *k = 0;
int i, n = ARRAYLEN(__keystroke.pool);
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
if (atomic_load_explicit(&__keystroke.pc, memory_order_acquire) < n &&
(i = atomic_fetch_add(&__keystroke.pc, 1)) < n) {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
k = __keystroke.pool + i;
} else {
Fix issue with pasting text into console
2023-09-21 20:28:52 +00:00
if ((e = dll_first(__keystroke.free))) {
k = KEYSTROKE_CONTAINER(e);
dll_remove(&__keystroke.free, &k->elem);
}
if (!k) {
if (_weaken(malloc)) {
k = _weaken(malloc)(sizeof(struct Keystroke));
} else {
enomem();
return 0;
}
}
}
if (k) {
bzero(k, sizeof(*k));
dll_init(&k->elem);
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
}
return k;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static textwindows void WriteTty(const char *p, size_t n) {
WriteFile(__keystroke.cot, p, n, 0, 0);
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
}
static textwindows bool IsCtl(int c) {
return isascii(c) && iscntrl(c) && c != '\n' && c != '\t';
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static textwindows void WriteCtl(const char *p, size_t n) {
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
size_t i;
for (i = 0; i < n; ++i) {
if (IsCtl(p[i])) {
char ctl[2];
ctl[0] = '^';
ctl[1] = p[i] ^ 0100;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
WriteTty(ctl, 2);
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
} else {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
WriteTty(p + i, 1);
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
}
}
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static textwindows void EchoTty(const char *p, size_t n) {
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
if (__ttyconf.magic & kTtyEchoRaw) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
WriteTty(p, n);
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
} else {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
WriteCtl(p, n);
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
}
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static textwindows void EraseCharacter(void) {
WriteTty("\b \b", 3);
}
static textwindows bool EraseKeystroke(void) {
Fix some additional Windows TTY issues
2023-10-04 06:53:59 +00:00
struct Dll *e;
if ((e = dll_last(__keystroke.line))) {
struct Keystroke *k = KEYSTROKE_CONTAINER(e);
dll_remove(&__keystroke.line, e);
dll_make_first(&__keystroke.free, e);
for (int i = k->buflen; i--;) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
if ((k->buf[i] & 0300) == 0200) continue; // utf-8 cont
EraseCharacter();
Fix some additional Windows TTY issues
2023-10-04 06:53:59 +00:00
if (!(__ttyconf.magic & kTtyEchoRaw) && IsCtl(k->buf[i])) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
EraseCharacter();
Fix some additional Windows TTY issues
2023-10-04 06:53:59 +00:00
}
}
return true;
} else {
return false;
}
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static textwindows void IngestConsoleInputRecord(struct NtInputRecord *r) {
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
// convert win32 console event into ansi
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
int len;
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
char buf[23];
if (!(len = ConvertConsoleInputToAnsi(r, buf))) {
return;
}
// handle backspace in canonical mode
if (len == 1 && buf[0] && //
(buf[0] & 255) == __ttyconf.verase && //
!(__ttyconf.magic & kTtyUncanon)) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
EraseKeystroke();
Fix some additional Windows TTY issues
2023-10-04 06:53:59 +00:00
return;
}
// handle kill in canonical mode
if (len == 1 && buf[0] && //
(buf[0] & 255) == __ttyconf.vkill && //
!(__ttyconf.magic & kTtyUncanon)) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
while (EraseKeystroke()) {
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
}
return;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// allocate object to hold keystroke
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
struct Keystroke *k;
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
if (!(k = NewKeystroke())) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
STRACE("out of keystroke memory");
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
return;
}
memcpy(k->buf, buf, sizeof(k->buf));
k->buflen = len;
// echo input if it was successfully recorded
// assuming the win32 console isn't doing it already
if (!(__ttyconf.magic & kTtySilence)) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
EchoTty(buf, len);
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
}
// save keystroke to appropriate list
if (__ttyconf.magic & kTtyUncanon) {
dll_make_last(&__keystroke.list, &k->elem);
} else {
dll_make_last(&__keystroke.line, &k->elem);
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// handle enter in canonical mode
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
if (len == 1 && buf[0] &&
((buf[0] & 255) == '\n' || //
(buf[0] & 255) == __ttyconf.veol || //
(buf[0] & 255) == __ttyconf.veol2)) {
dll_make_last(&__keystroke.list, __keystroke.line);
__keystroke.line = 0;
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
}
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
}
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static textwindows void IngestConsoleInput(void) {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
uint32_t i, n;
struct NtInputRecord records[16];
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
for (;;) {
if (__keystroke.end_of_file) return;
if (!GetNumberOfConsoleInputEvents(__keystroke.cin, &n)) {
goto UnexpectedEof;
}
if (!n) return;
n = MIN(ARRAYLEN(records), n);
if (!ReadConsoleInput(__keystroke.cin, records, n, &n)) {
goto UnexpectedEof;
}
for (i = 0; i < n && !__keystroke.end_of_file; ++i) {
IngestConsoleInputRecord(records + i);
}
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
UnexpectedEof:
STRACE("console read error %d", GetLastError());
__keystroke.end_of_file = true;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// Discards all unread stdin bytes.
textwindows int FlushConsoleInputBytes(void) {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
int rc;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
BLOCK_SIGNALS;
InitConsole();
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
LockKeystrokes();
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
FlushConsoleInputBuffer(__keystroke.cin);
dll_make_first(&__keystroke.free, __keystroke.list);
__keystroke.list = 0;
dll_make_first(&__keystroke.free, __keystroke.line);
__keystroke.line = 0;
rc = 0;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
UnlockKeystrokes();
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
ALLOW_SIGNALS;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
return rc;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// Returns number of stdin bytes that may be read without blocking.
textwindows int CountConsoleInputBytes(void) {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
struct Dll *e;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
int count = 0;
BLOCK_SIGNALS;
InitConsole();
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
LockKeystrokes();
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
IngestConsoleInput();
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
for (e = dll_first(__keystroke.list); e; e = dll_next(__keystroke.list, e)) {
count += KEYSTROKE_CONTAINER(e)->buflen;
}
if (!count && __keystroke.end_of_file) {
count = -1;
}
UnlockKeystrokes();
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
ALLOW_SIGNALS;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
return count;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// Intercept ANSI TTY commands that enable features.
textwindows void InterceptTerminalCommands(const char *data, size_t size) {
int i;
unsigned x;
bool ismouse;
uint32_t cm, cm2;
enum { ASC, ESC, CSI, CMD } t;
GetConsoleMode(GetConsoleInputHandle(), &cm), cm2 = cm;
for (ismouse = false, x = i = t = 0; i < size; ++i) {
switch (t) {
case ASC:
if (data[i] == 033) {
t = ESC;
}
break;
case ESC:
if (data[i] == '[') {
t = CSI;
} else {
t = ASC;
}
break;
case CSI:
if (data[i] == '?') {
t = CMD;
x = 0;
} else {
t = ASC;
}
break;
case CMD:
if ('0' <= data[i] && data[i] <= '9') {
x *= 10;
x += data[i] - '0';
} else if (data[i] == ';') {
ismouse |= IsMouseModeCommand(x);
x = 0;
} else if (data[i] == 'h') {
if (x == 1) {
__keystroke.vkt = kDecckm; // \e[?1h decckm on
Make some quick fixes
2023-10-09 00:56:59 +00:00
__keystroke.ohno_decckm = true;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
} else if ((ismouse |= IsMouseModeCommand(x))) {
__ttyconf.magic |= kTtyXtMouse;
cm2 |= kNtEnableMouseInput;
cm2 &= ~kNtEnableQuickEditMode; // take mouse
}
t = ASC;
} else if (data[i] == 'l') {
if (x == 1) {
__keystroke.vkt = kVirtualKey; // \e[?1l decckm off
Make some quick fixes
2023-10-09 00:56:59 +00:00
__keystroke.ohno_decckm = false;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
} else if ((ismouse |= IsMouseModeCommand(x))) {
__ttyconf.magic &= ~kTtyXtMouse;
cm2 |= kNtEnableQuickEditMode; // release mouse
}
t = ASC;
}
break;
default:
__builtin_unreachable();
}
}
if (cm2 != cm) {
SetConsoleMode(GetConsoleInputHandle(), cm2);
}
}
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
static textwindows bool DigestConsoleInput(char *data, size_t size, int *rc) {
// handle eof once available input is consumed
if (dll_is_empty(__keystroke.list) && __keystroke.end_of_file) {
*rc = 0;
return true;
}
// copy keystroke(s) into user buffer
int toto = 0;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
struct Dll *e;
Fix some additional Windows TTY issues
2023-10-04 06:53:59 +00:00
while (size && (e = dll_first(__keystroke.list))) {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
struct Keystroke *k = KEYSTROKE_CONTAINER(e);
uint32_t got = MIN(size, k->buflen);
uint32_t remain = k->buflen - got;
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
if (got) {
memcpy(data, k->buf, got);
data += got;
size -= got;
toto += got;
}
if (remain) {
memmove(k->buf, k->buf + got, remain);
} else {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
dll_remove(&__keystroke.list, e);
dll_make_first(&__keystroke.free, e);
}
k->buflen = remain;
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
if ((__ttyconf.magic & kTtyUncanon) && toto >= __ttyconf.vmin) {
break;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
}
}
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
// return result
if (toto) {
*rc = toto;
return true;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
} else {
Implement basic canonical mode for Windows The `cat` command now works properly, when run by itself on the bash command prompt. It's working beautifully so far, and is only missing a few keystrokes for clearing words and lines. Definitely works more well than the one that ships with WIN32 :-)
2023-10-04 05:34:45 +00:00
return false;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
}
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static textwindows int WaitForConsole(struct Fd *f, sigset_t waitmask) {
int rc;
sigset_t m;
int64_t sem;
uint32_t ms = -1u;
Make some quick fixes
2023-10-09 00:56:59 +00:00
struct PosixThread *pt;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
if (!__ttyconf.vmin) {
if (!__ttyconf.vtime) {
return 0; // non-blocking w/o raising eagain
} else {
ms = __ttyconf.vtime * 100;
}
}
if (f->flags & O_NONBLOCK) {
return eagain(); // standard unix non-blocking
}
Make some quick fixes
2023-10-09 00:56:59 +00:00
pt = _pthread_self();
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
pt->pt_flags |= PT_RESTARTABLE;
pt->pt_semaphore = sem = CreateSemaphore(0, 0, 1, 0);
pthread_cleanup_push((void *)CloseHandle, (void *)sem);
atomic_store_explicit(&pt->pt_blocker, PT_BLOCKER_SEM, memory_order_release);
m = __sig_beginwait(waitmask);
if ((rc = _check_cancel()) != -1 && (rc = _check_signal(true)) != -1) {
WaitForMultipleObjects(2, (int64_t[2]){sem, __keystroke.cin}, 0, ms);
if (~pt->pt_flags & PT_RESTARTABLE) rc = eintr();
if (rc == -1 && errno == EINTR) _check_cancel();
}
__sig_finishwait(m);
atomic_store_explicit(&pt->pt_blocker, PT_BLOCKER_CPU, memory_order_release);
pt->pt_flags &= ~PT_RESTARTABLE;
Make some quick fixes
2023-10-09 00:56:59 +00:00
pthread_cleanup_pop(true);
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
return rc;
}
static textwindows ssize_t ReadFromConsole(struct Fd *f, void *data,
size_t size, sigset_t waitmask) {
int rc;
bool done = false;
InitConsole();
do {
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
LockKeystrokes();
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
IngestConsoleInput();
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
done = DigestConsoleInput(data, size, &rc);
UnlockKeystrokes();
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
} while (!done && !(rc = WaitForConsole(f, waitmask)));
Rewrite Windows console input handling This change removes our use of ENABLE_VIRTUAL_TERMINAL_INPUT (which isn't very good) in favor of having read() translate Windows Console input events to ANSI/XTERM sequences by hand. This makes it possible to capture important keystrokes (e.g. ctrl-space) that weren't possible before. Most importantly this change also removes the stdin/sigwinch worker threads, which never really worked that well. Interactive TTY sessions will now work reliably when a Cosmo process spawns or forks another Cosmo process, e.g. unbourne.com launching emacs.com.
2023-09-19 18:42:38 +00:00
return rc;
}
Make the Windows Console work better The stdio reader thread now appears to be working recursively along cosmopolitan subprocesses. For example, it's now possible to launch vim.com from the unbourne.com bestline repl, thanks to hacks plus a bug fix to select() timeouts.
2023-09-07 23:03:19 +00:00
textwindows ssize_t sys_read_nt_impl(int fd, void *data, size_t size,
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
int64_t offset, sigset_t waitmask) {
Make improvements for Actually Portable Emacs - Get SIGWINCH working again on the New Technology - Correctly handle O_NOFOLLOW in open() on Windows - Implement synthetic umask() functionality on Windows - Do a better job managing file execute access on Windows - Fill in `st_uid` and `st_gid` with username hash on Windows - Munge UNICODE control pictures into control codes on Windows - Do a better job ensuring Windows console settings are restored - Introduce KPRINTF_LOG environment variable to log kprintf to a file
2023-08-19 13:41:06 +00:00
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// switch to terminal polyfill if reading from win32 console
struct Fd *f = g_fds.p + fd;
if (f->kind == kFdConsole) {
return ReadFromConsole(f, data, size, waitmask);
Fix writev() on the New Technology (#117)
2021-03-09 19:21:08 +00:00
}
Make important improvements - Fix preadv() and pwritev() for old distros - Introduce _npassert() and _unassert() macros - Prove that file locks work properly on Windows - Support fcntl(F_DUPFD_CLOEXEC) on more systems
2022-09-15 04:29:50 +00:00
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// perform heavy lifting
ssize_t rc;
rc = sys_readwrite_nt(fd, data, size, offset, f->handle, waitmask, ReadFile);
if (rc != -2) return rc;
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// mops up win32 errors
Fix bugs and make improvements - Get clone() working on FreeBSD - Increase some Python build quotas - Add more atomic builtins to chibicc - Fix ASAN poisoning of alloca() memory - Make MODE= mandatory link path tinier - Improve the examples folder a little bit - Start working on some more resource limits - Make the linenoise auto-complete UI as good as GNU readline - Update compile.com, avoiding AVX codegen on non-AVX systems - Make sure empty path to syscalls like opendir raises ENOENT - Correctly polyfill ENOENT vs. ENOTDIR on the New Technology - Port bestline's paredit features to //third_party/linenoise - Remove workarounds for RHEL 5.0 bugs that were fixed in 5.1
2022-04-20 16:56:53 +00:00
switch (GetLastError()) {
case kNtErrorBrokenPipe: // broken pipe
case kNtErrorNoData: // closing named pipe
case kNtErrorHandleEof: // pread read past EOF
return 0; //
case kNtErrorAccessDenied: // read doesn't return EACCESS
return ebadf(); //
default:
return __winerr();
}
Fix writev() on the New Technology (#117)
2021-03-09 19:21:08 +00:00
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static textwindows ssize_t sys_read_nt2(int fd, const struct iovec *iov,
size_t iovlen, int64_t opt_offset,
sigset_t waitmask) {
Fix writev() on the New Technology (#117)
2021-03-09 19:21:08 +00:00
ssize_t rc;
Apply some touchups
2021-02-07 14:11:44 +00:00
size_t i, total;
Make important improvements - Fix preadv() and pwritev() for old distros - Introduce _npassert() and _unassert() macros - Prove that file locks work properly on Windows - Support fcntl(F_DUPFD_CLOEXEC) on more systems
2022-09-15 04:29:50 +00:00
if (opt_offset < -1) return einval();
Initial import
2020-06-15 14:18:57 +00:00
while (iovlen && !iov[0].iov_len) iov++, iovlen--;
Apply some touchups
2021-02-07 14:11:44 +00:00
if (iovlen) {
for (total = i = 0; i < iovlen; ++i) {
Fix writev() on the New Technology (#117)
2021-03-09 19:21:08 +00:00
if (!iov[i].iov_len) continue;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
rc = sys_read_nt_impl(fd, iov[i].iov_base, iov[i].iov_len, opt_offset,
waitmask);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (rc == -1) {
if (total && errno != ECANCELED) {
return total;
} else {
return -1;
}
}
Fix writev() on the New Technology (#117)
2021-03-09 19:21:08 +00:00
total += rc;
if (opt_offset != -1) opt_offset += rc;
if (rc < iov[i].iov_len) break;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
waitmask = -1; // disable eintr/ecanceled for remaining iovecs
Apply some touchups
2021-02-07 14:11:44 +00:00
}
return total;
Initial import
2020-06-15 14:18:57 +00:00
} else {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
return sys_read_nt_impl(fd, NULL, 0, opt_offset, waitmask);
Initial import
2020-06-15 14:18:57 +00:00
}
}
Get Fat Emacs working in Windows Console
2023-08-18 11:55:57 +00:00
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
textwindows ssize_t sys_read_nt(int fd, const struct iovec *iov, size_t iovlen,
int64_t opt_offset) {
ssize_t rc;
sigset_t m = __sig_block();
rc = sys_read_nt2(fd, iov, iovlen, opt_offset, m);
__sig_unblock(m);
return rc;
}
Get Fat Emacs working in Windows Console
2023-08-18 11:55:57 +00:00
#endif /* __x86_64__ */
Reference in a new issue Copy permalink