build : enable more non-default compiler warnings (#3200)

examples/baby-llama/baby-llama.cpp

@@ -1,9 +1,12 @@
 #include "ggml.h"
 #include "train.h"
+
 #include <vector>
 #include <cassert>
-#include <random>
+#include <cstdlib>
 #include <cstring>
+#include <random>
+#include <vector>
 
 #if defined(_MSC_VER)
 #pragma warning(disable: 4244 4267) // possible loss of data
@@ -64,7 +67,7 @@ static struct ggml_tensor * randomize_tensor(
             break;
         default:
             assert(false);
-    };
+    }
 
     return tensor;
 }
@@ -389,7 +392,7 @@ static void randomize_model_lora(
     free_random_normal_distribution(rnd);
 }
 
-static bool init_kv_cache(struct llama_kv_cache* cache, struct llama_model * model, int n_batch) {
+static void init_kv_cache(struct llama_kv_cache* cache, struct llama_model * model, int n_batch) {
     const auto & hparams = model->hparams;
 
     const uint32_t n_ctx   = hparams.n_ctx;
@@ -415,14 +418,12 @@ static bool init_kv_cache(struct llama_kv_cache* cache, struct llama_model * mod
 
         if (!cache->ctx) {
             fprintf(stderr, "%s: failed to allocate memory for kv cache\n", __func__);
-            return false;
+            exit(1);
         }
     }
 
     cache->k = ggml_new_tensor_1d(cache->ctx, GGML_TYPE_F32, n_elements);
     cache->v = ggml_new_tensor_1d(cache->ctx, GGML_TYPE_F32, n_elements);
-
-    return true;
 }
 
 static bool init_kv_cache_lora(struct llama_kv_cache* cache, struct llama_model_lora * model, int n_batch) {

examples/llama-bench/llama-bench.cpp

@@ -655,9 +655,9 @@ struct printer {
     virtual ~printer() {}
 
     FILE * fout;
-    virtual void print_header(const cmd_params & params) { (void) params; };
+    virtual void print_header(const cmd_params & params) { (void) params; }
     virtual void print_test(const test & t) = 0;
-    virtual void print_footer() { };
+    virtual void print_footer() { }
 };
 
 struct csv_printer : public printer {

examples/main/main.cpp

@@ -852,7 +852,7 @@ int main(int argc, char ** argv) {
     llama_backend_free();
 
 #ifndef LOG_DISABLE_LOGS
-    LOG_TEE("Log end\n")
+    LOG_TEE("Log end\n");
 #endif // LOG_DISABLE_LOGS
 
     return 0;

examples/quantize/quantize.cpp

@@ -72,6 +72,7 @@ static bool try_parse_ftype(const std::string & ftype_str_in, llama_ftype & ftyp
 // usage:
 //  ./quantize [--allow-requantize] [--leave-output-tensor] models/llama/ggml-model.gguf [models/llama/ggml-model-quant.gguf] type [nthreads]
 //
+[[noreturn]]
 static void usage(const char * executable) {
     printf("usage: %s [--help] [--allow-requantize] [--leave-output-tensor] model-f32.gguf [model-quant.gguf] type [nthreads]\n\n", executable);
     printf("  --allow-requantize: Allows requantizing tensors that have already been quantized. Warning: This can severely reduce quality compared to quantizing from 16bit or 32bit\n");

examples/train-text-from-scratch/train-text-from-scratch.cpp

@@ -483,7 +483,7 @@ static struct ggml_tensor * llama_build_train_graphs(
 }
 
 #define GGUF_GET_KEY(ctx, dst, func, type, req, key) \
-{ \
+do { \
     const std::string skey(key); \
     const int kid = gguf_find_key(ctx, skey.c_str()); \
     if (kid >= 0) { \
@@ -495,7 +495,7 @@ static struct ggml_tensor * llama_build_train_graphs(
     } else if (req) { \
         die_fmt("key not found in model: %s", skey.c_str()); \
     } \
-}
+} while (0)
 
 static void load_llama_model_gguf(struct gguf_context * fctx, struct ggml_context * f_ggml_ctx, struct my_llama_model * model) {
     // NOTE: gguf_context must be initialized with f_ggml_ctx and no_alloc=false, otherwise tensor data can not be read
@@ -786,7 +786,7 @@ struct train_params {
     float rope_freq_scale;
 };
 
-struct train_params get_default_train_params() {
+static struct train_params get_default_train_params() {
     struct train_params params;
     params.common = get_default_train_params_common();
     params.fn_vocab_model    = "ggml-vic7b-uncensored-q4_0.bin";