Ivan Maidanski [Sat, 20 Jun 2015 19:22:07 +0000 (22:22 +0300)]
Adjust places where profiling callbacks invoked (to match Mono GC)
* alloc.c (GC_try_to_collect_inner): Move sending of GC_EVENT_START to
the beginning of function (i.e. send right after GC_dont_gc check);
move sending of GC_EVENT_END from GC_finish_collection to the end of
this function (send it only if collection completed successfully); add
TODO note about GC_EVENT_ABANDON notification.
* alloc.c (start_world_inner): Remove.
* alloc.c (GC_stopped_mark): Send GC_EVENT_PRE_STOP_WORLD,
GC_EVENT_POST_STOP_WORLD, GC_EVENT_PRE_START_WORLD,
GC_EVENT_POST_START_WORLD only if THREADS.
* alloc.c (GC_stopped_mark): Send GC_EVENT_MARK_START before
minimizing junk left in registers/stack (instead of after).
* alloc.c (GC_stopped_mark): Do not send GC_EVENT_MARK_END in case of
abandoned collection (add TODO note about GC_EVENT_MARK_ABANDON).
* alloc.c (GC_stopped_mark): Send GC_EVENT_MARK_END after checking
debugged objects for consistency (instead of before it).
* darwin_stop_world.c (GC_suspend_thread_list): Send
GC_EVENT_THREAD_SUSPENDED (in addition to that in GC_stop_world).
* darwin_stop_world.c (GC_thread_resume): Move sending of
GC_EVENT_THREAD_UNSUSPENDED from GC_start_world().
* pthread_stop_world.c (GC_suspend_all, GC_start_world): Do not send
GC_EVENT_THREAD_SUSPENDED if pthread_kill/android_thread_kill failed.
* pthread_stop_world.c (GC_start_world): Send
GC_EVENT_THREAD_UNSUSPENDED after pthread_resume_np (in case of
GC_OPENBSD_UTHREADS).
* win32_threads.c (GC_stop_world): Move sending of
GC_EVENT_THREAD_SUSPENDED to GC_suspend().
* win32_threads.c (GC_suspend, GC_start_world): Remove redundant cast
to void* of THREAD_HANDLE().
Ivan Maidanski [Sun, 17 May 2015 14:37:51 +0000 (17:37 +0300)]
Fix logic/compile errors in "event callbacks" code (Darwin, Pthreads)
* darwin_stop_world.c (GC_stop_world): Replace
GC_EVENT_THREAD_UNSUSPENDED to GC_EVENT_THREAD_SUSPENDED.
* pthread_stop_world.c (GC_on_collection_event): Declare.
* pthread_stop_world.c (GC_suspend_all): Change type of "thread_id"
local variable to pthread_t (or pid_t for Android) instead of int; fix
"threadid" parameter to "thread_id" in GC_on_collection_event call;
implement notification for NaCl target.
* pthread_stop_world.c (GC_start_world): Declare "thread_id" local
variable; implement notification for NaCl target.
Ivan Maidanski [Mon, 4 May 2015 20:21:06 +0000 (23:21 +0300)]
Avoid new[] debug variant definition for ancient MS VC++
(restore behavior broken in commit 3d784ed)
* include/gc_cpp.h (operator new[]): Update comment.
* include/gc_cpp.h (operator new[]): Do not define the debug variant
for MS VC pre-v7; add comment.
Ivan Maidanski [Fri, 8 May 2015 06:26:36 +0000 (09:26 +0300)]
Enable gc.h inclusion by client without implicit include windows.h (Win32)
* doc/README.macros (GC_DONT_INCLUDE_WINDOWS_H): Document.
* include/gc.h: Include windows.h unless GC_DONT_INCLUDE_WINDOWS_H
(only for Win32).
* include/gc.h (DECLSPEC_NORETURN): Define to __declspec(noreturn) by
default if windows.h not included (instead of empty).
* include/gc.h (GC_WIN32_SIZE_T): Define to DWORD (or unsigned long)
unless _WIN64 (instead of GC_uintptr_t which is unsigned int for
32-bit target).
* include/gc.h (GC_DllMain, GC_CreateThread, GC_ExitThread): If WINAPI
undefined (i.e., if windows.h not included by or prior to gc.h) then
declare the prototype using built-in C types instead of Windows types.
Ivan Maidanski [Thu, 16 Apr 2015 07:33:16 +0000 (10:33 +0300)]
Adjust code indentation of malloc/calloc/str[n]dup
(and do not define GC_init_lib_bounds if unused)
* malloc.c (GC_debug_malloc_replacement): Do not define/undefine macro
if REDIRECT_MALLOC_IN_HEADER.
* malloc.c (GC_libpthread_start, GC_libpthread_start,
GC_libpthread_end, GC_libld_start, GC_init_lib_bounds): Do not define
if REDIRECT_MALLOC_IN_HEADER.
* malloc.c (GC_debug_malloc_replacement, malloc, GC_init_lib_bounds,
calloc, strdup, strndup, free): Reformat code (adjust indentation) and
comments.
* malloc (free): Do not call GC_free if IGNORE_FREE defined.
Ivan Maidanski [Sun, 25 Nov 2012 08:07:13 +0000 (12:07 +0400)]
Recognize REDIRECT_MALLOC_IN_HEADER macro
* include/private/gcconfig.h: Do not issue error on REDIRECT_MALLOC
and THREADS if REDIRECT_MALLOC_IN_HEADER defined.
* malloc.c (malloc, calloc, strdup, strndup, free): Do not define if
REDIRECT_MALLOC_IN_HEADER.
* mallocx.c (realloc): Likewise.
Ivan Maidanski [Sun, 5 Apr 2015 15:59:45 +0000 (18:59 +0300)]
Do not pass VC-specific 'pragma message' in gc.h to other compilers (Win64)
* include/gc.h: Pass pragma message (to identify missing predefined
symbols for Win64) only if _MSC_VER (as it is specific to VC); report
the message only on first gc.h inclusion; adjust related comment.
Ivan Maidanski [Sat, 2 Aug 2014 08:17:57 +0000 (12:17 +0400)]
Resolve MessageBoxA at run-time (if requested) in GC_init (Win32)
(and code refactoring of GC_abort)
* misc.c (GC_win32_MessageBoxA): New static routine (only if MSWIN32,
move code from GC_abort).
* misc.c (GC_init, GC_abort): Replace MessageBoxA with
GC_win32_MessageBoxA call (to resolve "MessageBoxA" at run-time if
DONT_USE_USER32_DLL).
Ivan Maidanski [Sat, 2 Aug 2014 07:51:16 +0000 (11:51 +0400)]
Do not show WoW64 warning message unless CHECK_NOT_WOW64 defined (Win32)
* misc.c (GC_init): Do not show warning message box about WoW64 unless
CHECK_NOT_WOW64 (and GC_WIN32_THREADS); use GetModuleHandle instead of
GetModuleHandleA; check GetModuleHandle result; refine comment;
reformat code.
Ivan Maidanski [Mon, 7 Jul 2014 07:12:31 +0000 (11:12 +0400)]
New macro (NO_WINMAIN_ENTRY) to prefer main() instead of WinMain (Win32)
* tests/test.c (WinMain): Do not define if NO_WINMAIN_ENTRY (define
main() instead).
* tests/test_cpp.cc (WinMain): Define (instead of main()) for MSWIN32
and WinCE (but not for MinGW and not if NO_WINMAIN_ENTRY).
Ivan Maidanski [Sun, 18 Nov 2012 08:32:34 +0000 (12:32 +0400)]
Enable huge_test for Win64 (and LLP64 target)
* tests/huge_test.c (GC_WORD_MAX): New macro
* tests/huge_test.c (main): Do not check for long has the same size as
pointer; use unsigned GC_WORD_MAX instead of LONG_MAX; use NULL
instead of 0 for pointers.
Ivan Maidanski [Mon, 6 Apr 2015 17:06:47 +0000 (20:06 +0300)]
Do not include sigcontext.h and asm/sigcontext.h
(these Linux headers define sigcontext which is no longer used as
replaced with ucontext_t in GC 7.0alpha1)
* os_dep.c: Do not include linux/version.h (do not test
LINUX_VERSION_CODE), asm/signal.h, sigcontext.h, asm/sigcontext.h
since struct sigcontext not used in GC v7+.
Ivan Maidanski [Tue, 17 Mar 2015 23:32:30 +0000 (02:32 +0300)]
GC_stack_range_for code refactoring for Darwin/arm64
* darwin_stop_world.c (GC_stack_range_for): Replace "__" register name
prefix with THREAD_FLD(); roll x[0..28] push statements in a loop (for
AARCH64).
Ivan Maidanski [Fri, 27 Feb 2015 18:32:22 +0000 (21:32 +0300)]
Revert "Move asm machine-dependent files to 'src' folder" (partly)
Asm files moved back to base folder to avoid build issues (reported
for NetBSD and Solaris on Sparc). Alternative way is to adjust
configure.ac and Makefile.am properly but it requires more efforts
(including testing).
Ivan Maidanski [Fri, 23 Jan 2015 08:20:53 +0000 (11:20 +0300)]
GC_stack_range_for code refactoring (Darwin/arm)
* darwin_stop_world.c (GC_stack_range_for): Roll GC_push_one() calls
to a loop (for ARM32 only).
* os_dep.c: Remove redundant include mach/thread_status.h as already
included from gc_priv.h (for DARWIN only).
Bruce Hoult [Sun, 4 Jan 2015 15:32:29 +0000 (04:32 +1300)]
Fix OSX issue with snprintf wrapper macro
(bug introduced in commit 7bef74b)
OS X for some reason has problems with defining snprintf as a macro
and including another macro expansion in its arguments.
* cord/tests/cordtest.c (GC_SNPRINTF_BUFSZ_ARG): Remove.
* cord/tests/cordtest.c (GC_SNPRINTF): Do not define if no snprintf()
available.
* cord/tests/cordtest.c (test_printf): If GC_SNPRINTF undefined then
use sprintf() instead.
Bruce Hoult [Sun, 4 Jan 2015 15:33:35 +0000 (04:33 +1300)]
Fix OSX issue with pthread_attr_setstacksize failure
(bug introduced in commit 7bef74b)
* tests/test.c (main): Set stack size to 1000 KiB instead of 1000000 bytes (to avoid pthread_attr_setstacksize failure with
"stacksize is not a multiple of the system page size" error on
Darwin)
Ivan Maidanski [Tue, 4 Nov 2014 14:55:18 +0000 (17:55 +0300)]
Avoid explicit use of machine-specific x_THREAD_STATE macros (Darwin)
(code refactoring)
* include/private/gc_priv.h (GC_MACH_THREAD_STATE,
GC_MACH_THREAD_STATE_COUNT): Remove explicit definition to
PPC_THREAD_STATE[_COUNT], ARM_THREAD_STATE[_COUNT] (as the correct
values are defined via MACHINE_THREAD_STATE[_COUNT] macro).
Ivan Maidanski [Tue, 4 Nov 2014 14:01:08 +0000 (17:01 +0300)]
Fix mistyped ARM_THREAD_STATE macro (Darwin/arm)
* include/private/gc_priv.h (GC_MACH_THREAD_STATE,
GC_MACH_THREAD_STATE_COUNT): Define to ARM_THREAD_STATE[_COUNT],
respectively, instead of mistyped ARM_MACHINE_THREAD_STATE[_COUNT]
(if DARWIN and ARM32).
Tsugutomo Enami [Sun, 2 Nov 2014 07:46:17 +0000 (10:46 +0300)]
Fix FirstDLOpenedLinkMap for case libgc not 1st dynamically linked (NetBSD)
Current GC_FirstDLOpenedLinkMap() for NetBSD calls dlinfo(RTLD_SELF,
RTLD_DI_LINKMAP, &lm) to find link_map. So it will find link_map of
libgc. With guile's case, libgc is link to libguile and libguile is
linked to the guile command, so libgc is not the first one in the
link_map chain. That is why, data section of libguile, where
scm_protects exists, is not added to GC root and GC_is_visible fails.
* dyn_load.c (GC_FirstDLOpenedLinkMap): Iterate over link_map
(provided by dlinfo(RTLD_SELF)) to return 2nd element instead of the
provided one which might not always belong to libgc (only for NETBSD
and defined RTLD_DI_LINKMAP).
Ivan Maidanski [Mon, 20 Oct 2014 20:54:28 +0000 (00:54 +0400)]
Fix missing error handling of pthread_attr_init/getstacksize
* misc.c (GC_init): Explicitly ignore returned value of
pthread_mutexattr_destroy.
* os_dep.c (GC_get_main_stack_base, GC_get_stack_base, GC_dirty_init):
Likewise.
* pthread_support.c (start_mark_threads, pthread_create): Likewise.
* tests/test.c (main): Likewise.
* win32_threads.c (start_mark_threads): Likewise.
* pthread_support.c (pthread_create): ABORT (with the appropriate
message) in case of pthread_attr_getstacksize or pthread_attr_init
failure.
* tests/test.c (main): Print error code and FAIL if pthread_attr_init
or pthread_attr_setstacksize failed (only if GC_PTHREADS).
Yusuke Suzuki [Wed, 1 Oct 2014 18:38:02 +0000 (03:38 +0900)]
Fix __alloc_size__ availability detection (Clang)
Since __clang_major__/__clang_minor__ etc. are vendor dependent values,
we cannot implement the feature detection based on it.
For example, Apple clang versioning is different from the FreeBSD clang.
(At this time, Apple clang version is "6.0 (clang-600.0.51)" and
__clang_major__ is 6.)
Instead of this, we can use the clang feature detection macro,
__has_attribute.
* include/gc_config_macros.h (GC_ATTR_ALLOC_SIZE): Replace predefined
__clang_major/minor__ testing with __has_attribute() one (in case of
clang).
Ivan Maidanski [Sat, 27 Sep 2014 15:30:06 +0000 (19:30 +0400)]
Fix missing error handling of pthreads_mutex_init and cond_wait
* include/private/darwin_semaphore.h (sem_init): Destroy sem->mutex if
sem->cond initialization failed.
* include/private/darwin_semaphore.h (sem_post): Ignore
pthread_mutex_unlock result in case of pthread_cond_signal.
* include/private/darwin_semaphore.h (sem_wait): Unlock mutex and
return error (-1) if pthread_cond_wait failed.
* include/private/darwin_semaphore.h (sem_init): If pshared then
return -1 (with the appropriate errno code set) instead of ABORT.
* include/private/darwin_semaphore.h (sem_init, sem_post, sem_wait):
Treat non-zero value returned by pthread functions as error (instead
of only negative values).
* include/private/darwin_semaphore.h (sem_init, sem_post, sem_wait):
Reformat code.
* misc.c (GC_init): Abort (with the appropriate message) if
pthread_mutex[attr]_init failed (SN_TARGET_PS3 case only).
* specific.c (GC_key_create_inner): If pthread_mutex_init failed then
return its error code.
Use the ARM_THREAD_STATE32 flavor in the call to thread_get_state() when
running on 64-bit iOS 7+. If we don't iOS will assume we pass it an
arm_unified_thread_state_t while we actually pass it an arm_thread_state_t
which is a lot smaller. Without this fix thread_get_state() will corrupt the
stack and the app will crash.
Modified darwin_stop_world.c to use an arm_unified_thread_state_t struct to
store a thread's state when compiling against the iOS 7/8 SDK. Without this
patch GC_stack_range_for() crashes when running a 32-bit app on iOS 64-bit.
The old code passed an arm_thread_state_t and ARM_THREAD_STATE to
thread_get_state(). ARM_THREAD_STATE is the same as ARM_UNIFIED_THREAD_STATE
on iOS 7/8 and thread_get_state() actually expects an
arm_unified_thread_state_t. On iOS 32-bit it looks like thread_get_state()
only touches the first bytes corresponding to the size of arm_thread_state_t
so no crash there. On iOS 64-bit however it seems thread_get_state() writes to
the full arm_unified_thread_state_t which meant it would overflow the stack
allocated struct passed to it and mess up other values on the stack leading to
a crash later on.
Ivan Maidanski [Sat, 19 Jul 2014 08:52:54 +0000 (12:52 +0400)]
Fix and code refactoring of lock elision workaround (Linux/x64)
* configure.ac (HAVE_LIBC_VERSION_H, HAVE_GNU_GET_LIBC_VERSION): Remove
(revert change in previous commit).
* include/private/gcconfig.h (GLIBC_2_19_TSX_BUG): New macro defined
for Linux/x86_64 (if Glibc used) to workaround a bug in Glibc lock
elision implementation.
* pthread_support.c: Move include of gnu/libc-version.h to gcconfig.h
(used to check whether lock elision workaround needed).
* misc.c (GC_init): Reformat code.
* pthread_support.c (mark_mutex): Initialize (to
PTHREAD_MUTEX_INITIALIZER) even lock elision workaround is needed
(revert change in previous commit).
* pthread_support.c (parse_version): New static function (defined only
if GLIBC_2_19_TSX_BUG).
* pthread_support.c (GC_setup_mark_lock): Use parse_version to check
target Glibc version properly; do not reinitialize mutex unless
workaround needed; call ABORT (with the appropriate message) in case
of a failure in pthread_mutexattr_init/settype, pthread_mutex_init.
Paul Bone [Wed, 25 Jun 2014 01:17:50 +0000 (11:17 +1000)]
Workaround Linux NTPL lock elision bug.
glibc 2.19 on Linux x86-64 platforms includes support for lock elision,
by using Intel's TSX support when it is available. Without modifying an
application this converts suitable critical sections that use mutex into
transactional memory critical sections. See http://lwn.net/Articles/534758/
If a problem occurs that means that transactional memory can't be used, such
as a system call or buffer overflow, the pthreads implementation will catch
this error and retry the critical section using a normal mutex.
I noticed that since upgrading glibc that programs using Boehm GC crash, one
of these crashes was an assertion that the owner field of a mutex was
invalid. The assertion was generated by the pthreads implementation.
I believe that there is a bug in glibc that when a mutex cannot be used
safely for transactions that some series of events causes it's owner field
to be set incorrectly (or cleared when it shouldn't be).
I've found that I can work around this problem by having Boehm GC use an
error checking mutex, which I believe doesn't use lock elision and in my
testing doesn't crash.
XXX: This work-around mostly works except for linking the feature detection
in configure.ac to the conditional compilation in pthread_support.c as there
isn't an obvious way to make it work for automake and Makefile.direct.
Could I have some help updating the build system please?
include/private/pthread_support.h:
pthread_support.c:
Define GC_setup_mark_lock() This procedure creates the lock specifying a
pthread_mutexattr_t structure. This is used to disable lock elision on
Linux with glibc 2.19 or greater.
configure.ac:
If we're using Linux then check for the gnu extensions required to
identify the version of glibc at runtime.
misc.c:
Call GC_setup_mark_lock() when initialising the collector.
Ivan Maidanski [Sat, 2 Aug 2014 08:56:27 +0000 (12:56 +0400)]
Fix Unicode Win32 API calls in cord de_win
* cord/tests/de_win.c (FullAppName): Remove global variable.
* cord/tests/de_win.c (szAppName): Turn into macro; use TEXT().
* cord/tests/de_win.c (de_error): Force calling ANSI version of
MessageBox; remove casts to LPSTR.
* cord/tests/de_win.c (WinMain): Use TEXT() for lpszMenuName field and
for CreateWindow agrument.
* cord/tests/de_win.c (WndProc): Use TEXT() for DialogBox argument;
force calling ANSI version of TextOut.
Ivan Maidanski [Sun, 13 Jul 2014 08:18:16 +0000 (12:18 +0400)]
Code refactoring of Emscripten platform support (single-threaded)
* alloc.c (min_bytes_allocd): Test STACK_NOT_SCANNED macro instead
of __EMSCRIPTEN__ (stack size to scan is zero if STACK_NOT_SCANNED).
* include/private/gcconfig.h (ALIGNMENT): Remove duplicate definition
for _EMSCRIPTEN__.
* include/private/gcconfig.h (STACK_NOT_SCANNED): New macro defined
for __EMSCRIPTEN__ target (in addition to OS_TYPE, CPP_WORDSZ,
ALIGNMENT, DATASTART, DATAEND).
* mach_dep.c (GC_push_regs): Test STACK_NOT_SCANNED macro instead
of __EMSCRIPTEN__ (push nothing if STACK_NOT_SCANNED).
* mark_rts.c (GC_push_roots): Test STACK_NOT_SCANNED macro instead
of __EMSCRIPTEN__ (do not call GC_push_regs_and_stack if
STACK_NOT_SCANNED); mark cold_gc_frame argument as potentially unused.
* misc.c (GC_clear_stack): Test STACK_NOT_SCANNED macro instead
of __EMSCRIPTEN__ (do not clear stack if STACK_NOT_SCANNED).
* misc.c (GC_clear_stack): Reformat code.
* cord/tests/cordtest.c (GC_SNPRINTF): Redirect to sprintf also in
case of __STRICT_ANSI__ (i.e., if -ansi compiler option specified).
* misc.c (GC_VSNPRINTF): Redirect to vsprintf also in case of
__STRICT_ANSI__.
Ivan Maidanski [Sat, 21 Jun 2014 17:06:27 +0000 (21:06 +0400)]
Fix 'cord' headers path in EMX_MAKEFILE and README.QUICK
* EMX_MAKEFILE (CORD_INCLUDE_FILES): Replace cord/*.h files with
include/*.h ones (all "cord" header files moved in gc4.8).
* EMX_MAKEFILE (cords): Do not copy cord/*.h files to "include"
folder.
* README.QUICK: Replace cord/cord.h with include/cord.h; move sentence
about cord.h to the relevant paragraph.
Ivan Maidanski [Wed, 18 Jun 2014 23:04:28 +0000 (03:04 +0400)]
Fix unresolved vsnprintf in misc.c and snprintf in cordtest (DJGPP, VC)
* cord/tests/cordtest.c (GC_SNPRINTF, GC_SNPRINTF_BUFSZ_ARG): New
macro to workaround snprintf() missing in DJGPP and MS VC.
* cord/tests/cordtest.c (test_printf): Replace snprintf() with
GC_SNPRINTF and GC_SNPRINTF_BUFSZ_ARG.
* misc.c (GC_VSNPRINTF): Test DJGPP instead of NO_VSNPRINTF; refine
comment.
Ivan Maidanski [Sun, 15 Jun 2014 07:39:12 +0000 (11:39 +0400)]
Fix signedness of char values passed to isspace, iscntrl, isxdigit
(eliminate 'array subscript has type char' GCC warning)
* cord/tests/de_win.c (WinMain, plain_chars, control_chars): Cast char
value to unsigned char when passed to isspace(), iscntrl().
* os_dep.c (GC_parse_map_entry): Change type of "start_start",
"end_start", "maj_dev_start", "p" local variables from char* to
unsgined char* (as isspace(), isxdigit() expect non-negative values);
insert casts between char* and unsigned char* where needed.
* os_dep.c (GC_parse_map_entry): Remove "endp" local variable (use "p"
variable directly instead).
Ivan Maidanski [Sat, 14 Jun 2014 16:30:42 +0000 (20:30 +0400)]
Eliminate warnings detected by Cppcheck in cord de[_win]
* cord/tests/de.c (prune_map): Remove redundant first check of map for
NULL (as "map" variable has already been dereferenced at the point of
the check).
* cord/tests/de.c (replace_line): Do not define "len" local variable
if not used; eliminate duplicate CORD_len(s) call.
* cord/tests/de_win.c (WndProc): Cast char_width and char_height
global variables to unsigned int in division of unsigned xpos, ypos
local variables.
Ivan Maidanski [Sat, 14 Jun 2014 16:11:54 +0000 (20:11 +0400)]
Fix 'redundant assignment to itself' Cppcheck warning in GC_scratch_alloc
* headers.c (GC_scratch_alloc): Do not reassign "bytes_to_get" local
variable to avoid compiler warning in case of ROUNDUP_PAGESIZE_IF_MMAP
is a no-op (i.e., assign bytes_to_get only once directly to a rounded
value).
projects / gc / log