Actually Portable Python is now outperforming the Python binaries
that come bundled with Linux distros, at things like HTTP serving.
You can now have a fully featured Python install in just one .com
file that runs on six operating systems and is about 10mb in size.
With tuning, the tiniest is ~1mb. We've got most of the libraries
working, including pysqlite, and the repl now feels very pleasant.
The things you can't do quite yet are: threads and shared objects
but that can happen in the future, if the community falls in love
with this project and wants to see it developed further. Changes:
- Add siginterrupt()
- Add sqlite3 to Python
- Add issymlink() helper
- Make GetZipCdir() faster
- Add tgamma() and finite()
- Add legacy function lutimes()
- Add readlink() and realpath()
- Use heap allocations when appropriate
- Reorganize Python into two-stage build
- Save Lua / Python shell history to dotfile
- Integrate Python Lib embedding into linkage
- Make isregularfile() and isdirectory() go faster
- Make Python shell auto-completion work perfectly
- Make crash reports work better if changed directory
- Fix Python+NT open() / access() flag overflow error
- Disable Python tests relating to \N{LONG NAME} syntax
- Have Python REPL copyright() show all notice embeddings
The biggest technical challenge at the moment is working around
when Python tries to be too clever about filenames.
Thanks to all the refactorings we now have the ability to enforce
reasonable limitations on the amount of resources any individual
compile or test can consume. Those limits are currently:
- `-C 8` seconds of 3.1ghz CPU time
- `-M 256mebibytes` of virtual memory
- `-F 100megabyte` limit on file size
Only one file currently needs to exceed these limits:
o/$(MODE)/third_party/python/Objects/unicodeobject.o: \
QUOTA += -C16 # overrides cpu limit to 16 seconds
This change introduces a new sizetol() function to LIBC_FMT for parsing
byte or bit size strings with Si unit suffixes. Functions like atoi()
have been rewritten too.
This change gets the Python codebase into a state where it conforms to
the conventions of this codebase. It's now possible to include headers
from Python, without worrying about ordering. Python has traditionally
solved that problem by "diamonding" everything in Python.h, but that's
problematic since it means any change to any Python header invalidates
all the build artifacts. Lastly it makes tooling not work. Since it is
hard to explain to Emacs when I press C-c C-h to add an import line it
shouldn't add the header that actually defines the symbol, and instead
do follow the nonstandard Python convention.
Progress has been made on letting Python load source code from the zip
executable structure via the standard C library APIs. System calss now
recognizes zip!FILENAME alternative URIs as equivalent to zip:FILENAME
since Python uses colon as its delimiter.
Some progress has been made on embedding the notice license terms into
the Python object code. This is easier said than done since Python has
an extremely complicated ownership story.
- Some termios APIs have been added
- Implement rewinddir() dirstream API
- GetCpuCount() API added to Cosmopolitan Libc
- More bugs in Cosmopolitan Libc have been fixed
- zipobj.com now has flags for mangling the path
- Fixed bug a priori with sendfile() on certain BSDs
- Polyfill F_DUPFD and F_DUPFD_CLOEXEC across platforms
- FIOCLEX / FIONCLEX now polyfilled for fast O_CLOEXEC changes
- APE now supports a hybrid solution to no-self-modify for builds
- Many BSD-only magnums added, e.g. O_SEARCH, O_SHLOCK, SF_NODISKIO
Building o//third_party/python now takes 5 seconds on my PC
This change works towards modifying Python to use runtime dispatching
when appropriate. For example, when loading the magnums in the socket
module, it's a good idea to check if the magnum is zero, because that
means the local system platform doesn't support it.
You can now say the following in your redbean Lua code:
status,headers,payload = Fetch("https://foo.example")
The following Lua APIs have been introduced:
- Fetch(str) → str,{str:str},str
- GetHttpReason(int) → str
- GetHttpReason(int) → str
- ProgramSslFetchVerify(bool)
- ProgramSslClientVerify(bool)
The following flags have been introduced:
- `-j` enables client SSL verification
- `-k` disables Fetch() SSL verification
- `-t INT` may now be passed a negative value for keepalive
Lua exceptions now invoke Cosmopolitan's garbage collector when
unwinding the stack. So it's now safe to use _gc() w/ Lua 𝔱𝔥𝔯𝔬𝔴
See #97
- Use nullness checks when calling weakly linked functions.
- Avoid typedef for reasons described in Linux Kernel style guide.
- Avoid enum in in Windows headers. Earlier in Cosmo's history all one
hundred files in libc/nt/enum/ used to be enums and it resulted in
gigabytes of DWARF data almost as large as everything else in the
codebase combined.
- Bitfields aren't our friends. They have frequent ABI breakages,
inconsistent arithmetic across compilers, and different endianness
between cpus. Compiler authors also haven't invested much roi into
making bit fields go fast so they produce poor assembly.
- Use memccpy() instead of strncpy() or snprintf() for length-bounded
copying of C strings. strncpy() is a misunderstood function and
snprintf() is awesome but memccpy() deserves more love.
Your redbean can now interoperate with clients that require TLS crypto.
This is accomplished using a protocol polyglot that lets us distinguish
between HTTP and HTTPS regardless of the port number. Certificates will
be generated automatically, if none are supplied by the user. Footprint
increases by only a few hundred kb so redbean in MODY=tiny is now 1.0mb
- Add lseek() polyfills for ZIP executable
- Automatically polyfill /tmp/FOO paths on NT
- Fix readdir() / ftw() / nftw() bugs on Windows
- Introduce -B flag for slower SSL that's stronger
- Remove mbedtls features Cosmopolitan doesn't need
- Have base64 decoder support the uri-safe alternative
- Remove Truncated HMAC because it's forbidden by the IETF
- Add all the mbedtls test suites and make them go 3x faster
- Support opendir() / readdir() / closedir() on ZIP executable
- Use Everest for ECDHE-ECDSA because it's so good it's so good
- Add tinier implementation of sha1 since it's not worth the rom
- Add chi-square monte-carlo mean correlation tests for getrandom()
- Source entropy on Windows from the proper interface everyone uses
We're continuing to outperform NGINX and other servers on raw message
throughput. Using SSL means that instead of 1,000,000 qps you can get
around 300,000 qps. However redbean isn't as fast as NGINX yet at SSL
handshakes, since redbean can do 2,627 per second and NGINX does 4.3k
Right now, the SSL UX story works best if you give your redbean a key
signing key since that can be easily generated by openssl using a one
liner then redbean will do all the things that are impossibly hard to
do like signing ecdsa and rsa certificates that'll work in chrome. We
should integrate the let's encrypt acme protocol in the future.
Live Demo: https://redbean.justine.lol/
Root Cert: https://redbean.justine.lol/redbean1.crt
- Better UBSAN error messages
- POSIX Advisory Locks polyfills
- Move redbean manual to /.help.txt
- System call memory safety in ASAN mode
- Character classification now does UNICODE
redbean improvements:
- Explicitly disable corking
- Simulate Python regex API for Lua
- Send warmup requests in main process on startup
- Add Class-A granular IPv4 network classification
- Add /statusz page so you can monitor your redbean's health
- Fix regressions on OpenBSD/NetBSD caused by recent changes
- Plug Authorization header into Lua GetUser and GetPass APIs
- Recognize X-Forwarded-{For,Host} from local reverse proxies
- Add many additional functions to redbean Lua server page API
- Report resource usage of child processes on `/` listing page
- Introduce `-a` flag for logging child process resource usage
- Introduce `-t MILLIS` flag and `ProgramTimeout(ms)` init API
- Introduce `-H "Header: value"` flag and `ProgramHeader(k,v)` API
Cosmopolitan Libc improvements:
- Make strerror() simpler
- Make inet_pton() not depend on sscanf()
- Fix OpenExecutable() which broke .data section earlier
- Fix stdio in cases where it overflows kernel tty buffer
- Fix bugs in crash reporting w/o .com.dbg binary present
- Add polyfills for SO_LINGER, SO_RCVTIMEO, and SO_SNDTIMEO
- Polyfill TCP_CORK on BSD and XNU using TCP_NOPUSH magnums
New netcat clone in examples/nc.c:
While testing some of the failure conditions for redbean, I noticed that
BusyBox's `nc` command is pretty busted, if you use it as an interactive
tool, rather than having it be part of a pipeline. Unfortunately this'll
only work on UNIX since Windows doesn't let us poll on stdio and sockets
at the same time because I don't think they want tools like this running
on their platform. So if you want forbidden fruit, it's here so enjoy it
The most exciting improvement is dynamic pages will soon be able to use
the executable itself as an object store. it required a heroic technique
for overcoming ETXTBSY restrictions which lets us open the executable in
read/write mode, which means (1) wa can restore the APE header, and (2)
we can potentially containerize redbean extension code so that modules
you download for your redbean online will only impact your redbean.
Here's a list of breaking changes to redbean:
- Remove /tool/net/ prefix from magic ZIP paths
- GetHeader() now returns NIL if header is absent
Here's a list of fixes and enhancements to redbean:
- Support 64-bit ZIP archives
- Record User-Agent header in logs
- Add twelve error handlers to accept()
- Display octal st_mode on listing page
- Show ZIP file comments on listing page
- Restore APE MZ header on redbean startup
- Track request count on redbean index page
- Report server uptime on redbean index page
- Don't bind server socket using SO_REUSEPORT
- Fix#151 where Lua LoadAsset() could free twice
- Report rusage accounting when workers exit w/ -vv
- Use ZIP iattr field as text/plain vs. binary hint
- Add ParseUrl() API for parsing things like a.href
- Add ParseParams() API for parsing HTTP POST bodies
- Add IsAcceptablePath() API for checking dots, etc.
- Add IsValidHttpToken() API for validating sane ASCII
- Add IsAcceptableHostPort() for validating HOST[:PORT]
- Send 400 response to HTTP/1.1 requests without a Host
- Send 403 response if ZIP or file isn't other readable
- Add virtual hosting that tries prepending Host to path
- Route requests based on Host in Request-URI if present
- Host routing will attempt to remove or add the www. prefix
- Sign-extend UNIX timestamps and don't adjust FileTime zone
Here's some of the improvements made to Cosmopolitan Libc:
- Fix ape.S indentation
- Improve consts.sh magnums
- Write pretty good URL parser
- Improve rusage accounting apis
- Bring mremap() closer to working
- Added ZIP APIs which will change
- Check for overflow in reallocarray()
- Remove overly fancy linkage in strerror()
- Fix GDB attach on crash w/ OpenBSD msyscall()
- Make sigqueue() portable to most UNIX distros
- Make integer serialization macros more elegant
- Bring back 34x tprecode8to16() performance boost
- Make malloc() more resilient to absurdly large sizes
- removed unneeded share parameter from pipe on nt
- socktpair(type | SOCK_CLOEXEC) is now polyfilled
- use textwindows for linker micro-optimization
- apologies for auto clang-format diff noise :(
- improve socketpair docstring
See #122
- Polyfill open() w/ O_CLOEXEC on RHEL5
- Remove old workaround from rmdir() on the New Technology
- preadv() and pwritev() are now smarter about demodernization
- preadv() and pwritev() are now available on the New Technology
- Polyfill ucontext_t on FreeBSD/OpenBSD/NetBSD
- Add tests confirming signals can edit CPU state
- Work towards supporting ZIP filesystem on bare metal
- Add more tinymath unit tests for POSIX conformance
- Add X87 and SSE status flags to crash report
- Fix some bugs in blinkenlights
- Fix llvm build breakage
Your Actually Portable Executables now contains a simple virtual memory
that works similarly to the Linux Kernel in the sense that it maps your
physical memory to negative addresses. This is needed to support mmap()
and malloc(). This functionality has zero code size impact. For example
the MODE=tiny LIFE.COM executable is still only 12KB in size.
The APE bootloader code has also been simplified to improve readibility
and further elevate the elegance by which we're able to support so many
platforms thereby enhancing verifiability so that we may engender trust
in this bootloading process.
You can now build Cosmopolitan with Clang:
make -j8 MODE=llvm
o/llvm/examples/hello.com
The assembler and linker code is now friendly to LLVM too.
So it's not needed to configure Clang to use binutils under
the hood. If you love LLVM then you can now use pure LLVM.
It turns out adding OpenBSD msyscall() origin verification broke the
--ftrace flag. The executable needs to issue raw syscalls while it's
rewriting itself. So they need to be in the same section, and that's
just plain simpler too.
- Support deterministic stacks on OpenBSD
- Support OpenBSD system call origin verification
- Fix overrun by one in chibicc string token allocator
- Get all chibicc tests passing under Address Sanitizer
This change enables Address Sanitizer systemically w/ `make MODE=dbg`.
Our version of Rust's `unsafe` keyword is named `noasan` which is used
for two functions that do aligned memory chunking, like `strcpy.c` and
we need to fix the tiny DEFLATE code, but that's it everything else is
fabulous you can have all the fischer price security blankets you need
Best of all is we're now able to use the ASAN data in Blinkenlights to
colorize the memory dumps. See the screenshot below of a test program:
https://justine.lol/blinkenlights/asan.png
Which is operating on float arrays stored on the stack, with red areas
indicating poisoned memory, and the green areas indicate valid memory.
We always favor calling functions like openat(), fstatat(), etc. because
Linux, XNU, FreeBSD, and OpenBSD all elected to support them, while some
systems like Android love them so much, that they stopped supporting the
old interfaces.
This change ensures that when dirfd is actually a dirfd and not AT_FDCWD
we'll do the right thing on Windows NT. We use an API that's been around
since Vista to accomplish that.
This change also adds exponential backoff to chdir() on Windows since it
seems almost as flaky on Windows 7 as the rmdir() function.
Justine says nay for the time being. Only XNU implements this. It's not
clear what ABI XNU is using but it's obviously not the one in the POSIX
documentation link below. Since all platforms implement vfork, it might
be better to empirically gauge the intersection of consensus which will
have better performance than this interface.
https://pubs.opengroup.org/onlinepubs/009696699/xrat/xsh_chap03.html
For the first time ever, all tests in this codebase now pass, when
run automatically on macos, freebsd, openbsd, rhel5, rhel7, alpine
and windows via the network using the runit and runitd build tools
- Fix vfork exec path etc.
- Add XNU opendir() support
- Add OpenBSD opendir() support
- Add Linux history to syscalls.sh
- Use copy_file_range on FreeBSD 13+
- Fix system calls with 7+ arguments
- Fix Windows with greater than 16 FDs
- Fix RUNIT.COM and RUNITD.COM flakiness
- Fix OpenBSD munmap() when files are mapped
- Fix long double so it's actually long on Windows
- Fix OpenBSD truncate() and ftruncate() thunk typo
- Let Windows fcntl() be used on socket files descriptors
- Fix Windows fstat() which had an accidental printf statement
- Fix RHEL5 CLOCK_MONOTONIC by not aliasing to CLOCK_MONOTONIC_RAW
This is wonderful. I never could have dreamed it would be possible
to get it working so well on so many platforms with tiny binaries.
Fixes#31Fixes#25Fixes#14
It turned out that the linker was doing the wrong with the amalgamation
library concerning weak stubs. A regression test has been added and new
binaries have been uploaded to https://justine.lol/cosmopolitan/
Ideally this should be fixed by building a tool that turns multiple .a
files into a single .a file with deduplication. As a workaround for now
the cosmopolitan.a build is restructured to not include LIBC_STUBS which
meant technical debt needed to be paid off where non-stub interfaces
were moved to LIBC_INTRIN and LIBC_NEXGEN32E.
Thank @PerfectProductions in #31 for the report!
A new rollup tool now exists for flattening out the headers in a way
that works better for our purposes than cpp. A lot of the API clutter
has been removed. APIs that aren't a sure thing in terms of general
recommendation are now marked internal.
There's now a smoke test for the amalgamation archive and gigantic
header file. So we can now guarantee you can use this project on the
easiest difficulty setting without the gigantic repository.
A website is being created, which is currently a work in progress:
https://justine.storage.googleapis.com/cosmopolitan/index.html
This is done without using Microsoft's internal APIs. MAP_PRIVATE
mappings are copied to the subprocess via a pipe, since Microsoft
doesn't want us to have proper COW pages. MAP_SHARED mappings are
remapped without needing to do any copying. Global variables need
copying along with the stack and the whole heap of anonymous mem.
This actually improves the reliability of the redbean http server
although one shouldn't expect 10k+ connections on a home computer
that isn't running software built to serve like Linux or FreeBSD.
blinkenlights now does a pretty good job emulating what happens when
binaries boot from BIOS into long mode. So it's been much easier to
debug the bare metal process and wrinkle out many issues.
This change includes many bug fixes, for the NT polyfills, strings,
memory, boot, and math libraries which were discovered by adding more
tools for recreational programming, such as PC emulation. Lemon has also
been vendored because it works so well at parsing languages.
- Emulator can now test the αcτµαlly pδrταblε εxεcµταblε bootloader
- Whipped up a webserver named redbean. It services 150k requests per
second on a single core. Bundling assets inside zip enables extremely
fast serving for two reasons. The first is that zip central directory
lookups go faster than stat() system calls. The second is that both
zip and gzip content-encoding use DEFLATE, therefore, compressed
responses can be served via the sendfile() system call which does an
in-kernel copy directly from the zip executable structure. Also note
that red bean zip executables can be deployed easily to all platforms,
since these native executables work on Linux, Mac, BSD, and Windows.
- Address sanitizer now works very well
I wanted a tiny scriptable meltdown proof way to run userspace programs
and visualize how program execution impacts memory. It helps to explain
how things like Actually Portable Executable works. It can show you how
the GCC generated code is going about manipulating matrices and more. I
didn't feel fully comfortable with Qemu and Bochs because I'm not smart
enough to understand them. I wanted something like gVisor but with much
stronger levels of assurances. I wanted a single binary that'll run, on
all major operating systems with an embedded GPL barrier ZIP filesystem
that is tiny enough to transpile to JavaScript and run in browsers too.
https://justine.storage.googleapis.com/emulator625.mp4
