merge from dev

doc/mimalloc-doc.h

@@ -808,7 +808,7 @@ library so all calls to the standard `malloc` interface are
 resolved to the _mimalloc_ library.
 
 - `env LD_PRELOAD=/usr/lib/libmimalloc.so myprogram` (on Linux, BSD, etc.)
-- `env DYLD_INSERT_LIBRARIES=usr/lib/libmimalloc.dylib myprogram` (On macOS)
+- `env DYLD_INSERT_LIBRARIES=/usr/lib/libmimalloc.dylib myprogram` (On macOS)
 
   Note certain security restrictions may apply when doing this from
   the [shell](https://stackoverflow.com/questions/43941322/dyld-insert-libraries-ignored-when-calling-application-through-bash).

docs/overrides.html

@@ -109,7 +109,7 @@ $(document).ready(function(){initNavTree('overrides.html','');});
 <p>On these systems we preload the mimalloc shared library so all calls to the standard <code>malloc</code> interface are resolved to the <em>mimalloc</em> library.</p>
 <ul>
 <li><code>env LD_PRELOAD=/usr/lib/libmimalloc.so myprogram</code> (on Linux, BSD, etc.)</li>
-<li><p class="startli"><code>env DYLD_INSERT_LIBRARIES=usr/lib/libmimalloc.dylib myprogram</code> (On macOS)</p>
+<li><p class="startli"><code>env DYLD_INSERT_LIBRARIES=/usr/lib/libmimalloc.dylib myprogram</code> (On macOS)</p>
 <p class="startli">Note certain security restrictions may apply when doing this from the <a href="https://stackoverflow.com/questions/43941322/dyld-insert-libraries-ignored-when-calling-application-through-bash">shell</a>.</p>
 </li>
 </ul>

include/mimalloc-internal.h

@@ -117,6 +117,9 @@ bool        _mi_page_is_valid(mi_page_t* page);
 #define mi_likely(x)       (x)
 #endif
 
+#ifndef __has_builtin
+#define __has_builtin(x)  0
+#endif
 
 #if defined(_MSC_VER)
 #define mi_decl_noinline   __declspec(noinline)
@@ -149,9 +152,17 @@ bool        _mi_page_is_valid(mi_page_t* page);
 // Overflow detecting multiply
 #define MI_MUL_NO_OVERFLOW ((size_t)1 << (4*sizeof(size_t)))  // sqrt(SIZE_MAX)
 static inline bool mi_mul_overflow(size_t size, size_t count, size_t* total) {
+#if __has_builtin(__builtin_umul_overflow) || __GNUC__ >= 5
+#if (MI_INTPTR_SIZE == 4)
+  return __builtin_umul_overflow(size, count, total);
+#else
+  return __builtin_umull_overflow(size, count, total);
+#endif
+#else /* __builtin_umul_overflow is unavailable */
   *total = size * count;
   return ((size >= MI_MUL_NO_OVERFLOW || count >= MI_MUL_NO_OVERFLOW)
           && size > 0 && (SIZE_MAX / size) < count);
+#endif
 }
 
 // Align a byte size to a size in _machine words_,

include/mimalloc-types.h

@@ -8,7 +8,6 @@ terms of the MIT license. A copy of the license can be found in the file
 #ifndef MIMALLOC_TYPES_H
 #define MIMALLOC_TYPES_H
 
-#include <stdlib.h>   // size_t etc.
 #include <stddef.h>   // ptrdiff_t
 #include <stdint.h>   // uintptr_t, uint16_t, etc
 

include/mimalloc.h

@@ -69,7 +69,6 @@ terms of the MIT license. A copy of the license can be found in the file
 // Includes
 // ------------------------------------------------------
 
-#include <stdlib.h>     // size_t, malloc etc.
 #include <stdbool.h>    // bool
 #include <stdio.h>      // FILE
 

readme.md

@@ -191,7 +191,7 @@ library so all calls to the standard `malloc` interface are
 resolved to the _mimalloc_ library.
 
 - `env LD_PRELOAD=/usr/lib/libmimalloc.so myprogram` (on Linux, BSD, etc.)
-- `env DYLD_INSERT_LIBRARIES=usr/lib/libmimalloc.dylib myprogram` (On macOS)
+- `env DYLD_INSERT_LIBRARIES=/usr/lib/libmimalloc.dylib myprogram` (On macOS)
 
   Note certain security restrictions may apply when doing this from
   the [shell](https://stackoverflow.com/questions/43941322/dyld-insert-libraries-ignored-when-calling-application-through-bash).

src/alloc-aligned.c

@@ -8,7 +8,7 @@ terms of the MIT license. A copy of the license can be found in the file
 #include "mimalloc.h"
 #include "mimalloc-internal.h"
 
-#include <string.h>  // memset
+#include <string.h>  // memset, memcpy
 
 // ------------------------------------------------------
 // Aligned Allocation

src/alloc-posix.c

@@ -19,6 +19,7 @@ terms of the MIT license. A copy of the license can be found in the file
 
 #include <errno.h>
 #include <string.h>  // memcpy
+#include <stdlib.h>  // getenv
 
 #ifndef EINVAL
 #define EINVAL 22
@@ -115,7 +116,7 @@ int mi_dupenv_s(char** buf, size_t* size, const char* name) mi_attr_noexcept {
   #pragma warning(suppress:4996)
   char* p = getenv(name);
   if (p==NULL) {
-    *buf = NULL;    
+    *buf = NULL;
   }
   else {
     *buf = mi_strdup(p);
@@ -146,4 +147,3 @@ int mi_wdupenv_s(unsigned short** buf, size_t* size, const unsigned short* name)
   return 0;
 #endif
 }
-

src/alloc.c

@@ -8,7 +8,8 @@ terms of the MIT license. A copy of the license can be found in the file
 #include "mimalloc-internal.h"
 #include "mimalloc-atomic.h"
 
-#include <string.h>  // memset
+#include <string.h>  // memset, memcpy, strlen
+#include <stdlib.h>  // malloc, exit
 
 #define MI_IN_ALLOC_C
 #include "alloc-override.c"
@@ -465,7 +466,7 @@ char* mi_heap_realpath(mi_heap_t* heap, const char* fname, char* resolved_name)
   }
 }
 #else
-#include <unistd.h>
+#include <unistd.h>  // pathconf
 static size_t mi_path_max() {
   static size_t path_max = 0;
   if (path_max <= 0) {

src/init.c

@@ -7,7 +7,8 @@ terms of the MIT license. A copy of the license can be found in the file
 #include "mimalloc.h"
 #include "mimalloc-internal.h"
 
-#include <string.h>  // memcpy
+#include <string.h>  // memcpy, memset
+#include <stdlib.h>  // atexit
 
 // Empty page used to initialize the small free pages array
 const mi_page_t _mi_page_empty = {

src/options.c

@@ -9,7 +9,8 @@ terms of the MIT license. A copy of the license can be found in the file
 #include "mimalloc-atomic.h"
 
 #include <stdio.h>
-#include <string.h> // strcmp
+#include <stdlib.h> // strtol
+#include <string.h> // strncpy, strncat, strlen, strstr
 #include <ctype.h>  // toupper
 #include <stdarg.h>
 
@@ -59,7 +60,7 @@ static mi_option_desc_t options[_mi_option_last] =
   { 0, UNINIT, "large_os_pages" },      // use large OS pages, use only with eager commit to prevent fragmentation of VMA's
   { 0, UNINIT, "page_reset" },
   { 0, UNINIT, "cache_reset" },
-  { 0, UNINIT, "reset_decommits" },     // note: cannot enable this if secure is on 
+  { 0, UNINIT, "reset_decommits" },     // note: cannot enable this if secure is on
   { 0, UNINIT, "reset_discards" }       // note: cannot enable this if secure is on
 };
 
@@ -68,7 +69,7 @@ static void mi_option_init(mi_option_desc_t* desc);
 long mi_option_get(mi_option_t option) {
   mi_assert(option >= 0 && option < _mi_option_last);
   mi_option_desc_t* desc = &options[option];
-  if (desc->init == UNINIT) {
+  if (mi_unlikely(desc->init == UNINIT)) {
     mi_option_init(desc);
     if (option != mi_option_verbose) {
       _mi_verbose_message("option '%s': %ld\n", desc->name, desc->value);
@@ -116,15 +117,15 @@ static void mi_vfprintf( FILE* out, const char* prefix, const char* fmt, va_list
   char buf[256];
   if (fmt==NULL) return;
   if (out==NULL) out = stdout;
-  if (_mi_preloading()) return;  
+  if (_mi_preloading()) return;
   vsnprintf(buf,sizeof(buf)-1,fmt,args);
   #ifdef _WIN32
-  // on windows with redirection, the C runtime uses us and we cannot call `fputs` 
+  // on windows with redirection, the C runtime uses us and we cannot call `fputs`
   // while called from the C runtime itself, so use a non-locking option
-  if (out==stderr) { 
-    if (prefix != NULL) _cputs(prefix); 
-    _cputs(buf); 
-    return; 
+  if (out==stderr) {
+    if (prefix != NULL) _cputs(prefix);
+    _cputs(buf);
+    return;
   }
   #endif
   if (prefix != NULL) fputs(prefix,out);
@@ -204,10 +205,12 @@ static const char* mi_getenv(const char* name) {
   const char* s = getenv(name);
   if (s == NULL) {
     char buf[64+1];
-    mi_strlcpy(buf,name,64);
-    for (size_t i = 0; i < strlen(buf); i++) {
+    size_t len = strlen(name);
+    if (len >= sizeof(buf)) len = sizeof(buf) - 1;
+    for (size_t i = 0; i < len; i++) {
       buf[i] = toupper(name[i]);
     }
+    buf[len] = 0;
     #pragma warning(suppress:4996)
     s = getenv(buf);
   }
@@ -217,15 +220,17 @@ static const char* mi_getenv(const char* name) {
 static void mi_option_init(mi_option_desc_t* desc) {
   if (!_mi_preloading()) desc->init = DEFAULTED;
   // Read option value from the environment
-  char buf[64];
+  char buf[64+1];
   mi_strlcpy(buf, "mimalloc_", sizeof(buf));
   mi_strlcat(buf, desc->name, sizeof(buf));
-  const char* s = mi_getenv(buf);  
+  const char* s = mi_getenv(buf);
   if (s != NULL) {
-    mi_strlcpy(buf, s, sizeof(buf));
-    for (size_t i = 0; i < strlen(buf); i++) {
-      buf[i] = toupper(buf[i]);
+    size_t len = strlen(s);
+    if (len >= sizeof(buf)) len = sizeof(buf) - 1;
+    for (size_t i = 0; i < len; i++) {
+      buf[i] = toupper(s[i]);
     }
+    buf[len] = 0;
     if (buf[0]==0 || strstr("1;TRUE;YES;ON", buf) != NULL) {
       desc->value = 1;
       desc->init = INITIALIZED;

src/os.c

@@ -12,7 +12,7 @@ terms of the MIT license. A copy of the license can be found in the file
 #include "mimalloc-internal.h"
 #include "mimalloc-atomic.h"
 
-#include <string.h>  // memset
+#include <string.h>  // strerror
 #include <errno.h>
 
 #if defined(_WIN32)
@@ -211,7 +211,7 @@ static void* mi_win_virtual_alloc(void* addr, size_t size, size_t try_alignment,
   if (use_large_os_page(size, try_alignment)) {
     uintptr_t try_ok = mi_atomic_read(&large_page_try_ok);
     if (try_ok > 0) {
-      // if a large page page allocation fails, it seems the calls to VirtualAlloc get very expensive.
+      // if a large page allocation fails, it seems the calls to VirtualAlloc get very expensive.
       // therefore, once a large page allocation failed, we don't try again for `large_page_try_ok` times.
       mi_atomic_compare_exchange(&large_page_try_ok, try_ok - 1, try_ok);
     }

src/page.c

@@ -15,8 +15,6 @@ terms of the MIT license. A copy of the license can be found in the file
 #include "mimalloc-internal.h"
 #include "mimalloc-atomic.h"
 
-#include <string.h>  // memset, memcpy
-
 /* -----------------------------------------------------------
   Definition of page queues for each block size
 ----------------------------------------------------------- */

test/test-stress.c

@@ -180,7 +180,7 @@ static DWORD WINAPI thread_entry(LPVOID param) {
 static void run_os_threads(size_t nthreads) {
   DWORD* tids = (DWORD*)malloc(nthreads * sizeof(DWORD));
   HANDLE* thandles = (HANDLE*)malloc(nthreads * sizeof(HANDLE));
-  for (intptr_t i = 0; i < nthreads; i++) {
+  for (uintptr_t i = 0; i < nthreads; i++) {
     thandles[i] = CreateThread(0, 4096, &thread_entry, (void*)(i), 0, &tids[i]);
   }
   for (int i = 0; i < nthreads; i++) {