merge from dev

src/alloc-override.c

@@ -248,7 +248,7 @@ extern "C" {
   // Forward Posix/Unix calls as well
   void*  reallocf(void* p, size_t newsize) MI_FORWARD2(mi_reallocf,p,newsize)
   size_t malloc_size(const void* p)        MI_FORWARD1(mi_usable_size,p)
-  #if !defined(__ANDROID__) && !defined(__FreeBSD__)
+  #if !defined(__ANDROID__) && !defined(__FreeBSD__) && !defined(__DragonFly__)
   size_t malloc_usable_size(void *p)       MI_FORWARD1(mi_usable_size,p)
   #else
   size_t malloc_usable_size(const void *p) MI_FORWARD1(mi_usable_size,p)

src/heap.c

@@ -390,9 +390,10 @@ void mi_heap_destroy(mi_heap_t* heap) {
 }
 
 // forcefully destroy all heaps in the current thread
-void _mi_heap_unsafe_destroy_all(void) {
-  mi_heap_t* bheap = mi_heap_get_backing();
-  mi_heap_t* curr = bheap->tld->heaps;
+void _mi_heap_unsafe_destroy_all(mi_heap_t* heap) {
+  mi_assert_internal(heap != NULL);
+  if (heap == NULL) return;
+  mi_heap_t* curr = heap->tld->heaps;
   while (curr != NULL) {
     mi_heap_t* next = curr->next;
     if (!curr->allow_page_reclaim) {

src/init.c

@@ -707,15 +707,20 @@ void mi_cdecl _mi_process_done(void) {
   if (process_done) return;
   process_done = true;
 
+  // get the default heap so we don't need to acces thread locals anymore
+  mi_heap_t* heap = mi_prim_get_default_heap();  // use prim to not initialize any heap
+  mi_assert_internal(heap != NULL);
+
   // release any thread specific resources and ensure _mi_thread_done is called on all but the main thread
   _mi_prim_thread_done_auto_done();
 
+
   #ifndef MI_SKIP_COLLECT_ON_EXIT
     #if (MI_DEBUG || !defined(MI_SHARED_LIB))
     // free all memory if possible on process exit. This is not needed for a stand-alone process
     // but should be done if mimalloc is statically linked into another shared library which
     // is repeatedly loaded/unloaded, see issue #281.
-    mi_collect(true /* force */ );
+    mi_heap_collect(heap, true /* force */ );
     #endif
   #endif
 
@@ -723,9 +728,9 @@ void mi_cdecl _mi_process_done(void) {
   // since after process_done there might still be other code running that calls `free` (like at_exit routines,
   // or C-runtime termination code.
   if (mi_option_is_enabled(mi_option_destroy_on_exit)) {
-    mi_collect(true /* force */);
-    _mi_heap_unsafe_destroy_all();     // forcefully release all memory held by all heaps (of this thread only!)
-    _mi_arenas_unsafe_destroy_all(&tld_main);
+    mi_heap_collect(heap, true /* force */);
+    _mi_heap_unsafe_destroy_all(heap);     // forcefully release all memory held by all heaps (of this thread only!)
+    _mi_arenas_unsafe_destroy_all(heap->tld);
   }
 
   if (mi_option_is_enabled(mi_option_show_stats) || mi_option_is_enabled(mi_option_verbose)) {

src/options.c

@@ -158,7 +158,7 @@ static mi_option_desc_t options[_mi_option_last] =
   { MI_DEFAULT_DISALLOW_ARENA_ALLOC,   UNINIT, MI_OPTION(disallow_arena_alloc) }, // 1 = do not use arena's for allocation (except if using specific arena id's)
   { 400, UNINIT, MI_OPTION(retry_on_oom) },             // windows only: retry on out-of-memory for N milli seconds (=400), set to 0 to disable retries.
 #if defined(MI_VISIT_ABANDONED)
-  { 1,   INITIALIZED, MI_OPTION(visit_abandoned) },     // allow visiting heap blocks in abandonded segments; requires taking locks during reclaim.
+  { 1,   INITIALIZED, MI_OPTION(visit_abandoned) },     // allow visiting heap blocks in abandoned segments; requires taking locks during reclaim.
 #else
   { 0,   UNINIT, MI_OPTION(visit_abandoned) },
 #endif
@@ -386,7 +386,7 @@ static _Atomic(size_t) warning_count; // = 0;  // when >= max_warning_count stop
 // (recursively) invoke malloc again to allocate space for the thread local
 // variables on demand. This is why we use a _mi_preloading test on such
 // platforms. However, C code generator may move the initial thread local address
-// load before the `if` and we therefore split it out in a separate funcion.
+// load before the `if` and we therefore split it out in a separate function.
 static mi_decl_thread bool recurse = false;
 
 static mi_decl_noinline bool mi_recurse_enter_prim(void) {

src/prim/unix/prim.c

@@ -27,6 +27,7 @@ terms of the MIT license. A copy of the license can be found in the file
 #include <sys/mman.h>  // mmap
 #include <unistd.h>    // sysconf
 #include <fcntl.h>     // open, close, read, access
+#include <stdlib.h>    // getenv, arc4random_buf
 
 #if defined(__linux__)
   #include <features.h>
@@ -248,7 +249,7 @@ static int unix_mmap_fd(void) {
   #if defined(VM_MAKE_TAG)
   // macOS: tracking anonymous page with a specific ID. (All up to 98 are taken officially but LLVM sanitizers had taken 99)
   int os_tag = (int)mi_option_get(mi_option_os_tag);
-  if (os_tag < 100 || os_tag > 255) { os_tag = 100; }
+  if (os_tag < 100 || os_tag > 255) { os_tag = 254; }
   return VM_MAKE_TAG(os_tag);
   #else
   return -1;
@@ -774,7 +775,7 @@ bool _mi_prim_getenv(const char* name, char* result, size_t result_size) {
 #include <CommonCrypto/CommonRandom.h>
 
 bool _mi_prim_random_buf(void* buf, size_t buf_len) {
-  // We prefere CCRandomGenerateBytes as it returns an error code while arc4random_buf
+  // We prefer CCRandomGenerateBytes as it returns an error code while arc4random_buf
   // may fail silently on macOS. See PR #390, and <https://opensource.apple.com/source/Libc/Libc-1439.40.11/gen/FreeBSD/arc4random.c.auto.html>
   return (CCRandomGenerateBytes(buf, buf_len) == kCCSuccess);
 }
@@ -784,7 +785,6 @@ bool _mi_prim_random_buf(void* buf, size_t buf_len) {
       defined(__sun) || \
       (defined(__APPLE__) && (MAC_OS_X_VERSION_MIN_REQUIRED >= MAC_OS_X_VERSION_10_7))
 
-#include <stdlib.h>
 bool _mi_prim_random_buf(void* buf, size_t buf_len) {
   arc4random_buf(buf, buf_len);
   return true;

src/prim/windows/prim.c

@@ -58,7 +58,7 @@ typedef NTSTATUS (__stdcall *PNtAllocateVirtualMemoryEx)(HANDLE, PVOID*, SIZE_T*
 static PVirtualAlloc2 pVirtualAlloc2 = NULL;
 static PNtAllocateVirtualMemoryEx pNtAllocateVirtualMemoryEx = NULL;
 
-// Similarly, GetNumaProcesorNodeEx is only supported since Windows 7
+// Similarly, GetNumaProcessorNodeEx is only supported since Windows 7
 typedef struct MI_PROCESSOR_NUMBER_S { WORD Group; BYTE Number; BYTE Reserved; } MI_PROCESSOR_NUMBER;
 
 typedef VOID (__stdcall *PGetCurrentProcessorNumberEx)(MI_PROCESSOR_NUMBER* ProcNumber);