Path to this page:
Subject: CVS commit: pkgsrc/devel/boehm-gc
From: Thomas Klausner
Date: 2023-10-19 17:36:26
Message id: 20231019153626.4AAE3FADC@cvs.NetBSD.org
Log Message:
boehm-gc: update to 8.2.4.
== [8.2.4] 2023-05-26 ==
* Abort with appropriate message if first call of mmap fails with EPERM
* Adjust CORD_ec comment placement in ec.h
* Adjust WoW64 workaround to work on UWP/WinRT
* Adjust naming of Win32/64 and x86/64 words in comments and documentation
* Avoid potential race between realloc and GC_block_was_dirty
* Do not double-clear first two words of object in GC_generic_malloc_aligned
* Do not mention FASTLOCK in comment
* Do not mix debug and non-debug allocations in disclaim tests
* Do not prohibit threaded builds with malloc redirection on non-Linux
* Do not prohibit zero proc argument in GC_register_disclaim_proc
* Eliminate '&array may not produce intended result' wcc warnings
* Eliminate 'GC_unmap_end declared but unused' bcc warning in disclaim_bench
* Eliminate 'ISO C++17 does not allow register specifier' gcc warning
* Eliminate 'cast signed to bigger unsigned' CSA warning in WARNs, new_thread
* Eliminate 'n obtained from untrusted source' code defect FP in test_cpp
* Eliminate 'skipping config since MAXSIG/_NSIG is unknown' cppcheck FP
* Eliminate data race FP between remove_protection and write_fault_handler
* Eliminate data race FP reported in need_unreachable_finalization by TSan
* Ensure 'new' system header is included by gc_cpp.h if GC_INCLUDE_NEW
* Ensure GC_NO_PTHREAD_SIGMASK defined if no GC_pthread_sigmask prototype
* Fix 'EMSCRIPTEN macro redefined' compile error
* Fix 'ISO C90 forbids mixed decl and code' warning in SOFT_VDB dirty_init
* Fix 'call to undeclared pthread_setname_np' errors in configure and cmake
* Fix 'failed to create new win32 semaphore' Cygwin fatal error at fork
* Fix 'operator new is missing throw(bad_alloc)' clang warning in gc_cpp.h
* Fix 'overflow in conversion from word' g++ warning in GC_init
* Fix 'unknown option --no-undefined' linker error in cmake script (OS X)
* Fix 'unresolved _end' linker error when targeting Android bitcode
* Fix CORD_next() indent inside loop in test_basics() of cordtest
* Fix DCL_LOCK_STATE placement in GC_set_oom_fn
* Fix GC_excl_table overrun on overflow in GC_exclude_static_roots
* Fix GC_thread_is_registered for finished threads
* Fix GC_unreachable_finalize_mark_proc to ensure its unique address
* Fix GC_unregister_my_thread call before GC functions usage in gctest
* Fix IRIX5 defined wrongly on FreeBSD/mips, WinCE/mips, Tandem S-Series
* Fix allocated objects count increment in alloc8bytes of gctest
* Fix alt-stack handling in GC_push_all_stacks if stack grows up
* Fix comparisons to heap boundary in GC_get_back_ptr_info and GC_mark_from
* Fix data race in GC_heapsize_at_forced_unmap variable
* Fix description of client promise for IGNORE_OFF_PAGE allocated objects
* Fix disabling of automatic dynamic libraries registration
* Fix double initialization of main thread local free lists on Win32
* Fix gccpp and gctba library names in gcinterface.md
* Fix infinite loop in disable_gc_for_dlopen and GC_wait_for_gc_completion
* Fix infinite wait in pthread_join/detach if thread already finished (Win32)
* Fix joinable threads shutdown on NaCl
* Fix loop condition over dll_thread_table in GC_lookup_pthread (Win32)
* Fix missing GC_CALLBACK for GC_waitForSingleObjectInfinite
* Fix missing extern C for __asan_default_options
* Fix missing libalphagc.so dependency in Makefile.direct
* Fix missing lock while updating GC_in_thread_creation in GC_exit_check
* Fix missing recovery from faults in GC_mark_some on Win64 if MinGW
* Fix missing result check of pthread_attr_getdetachstate in pthread_create
* Fix mistyped function name in documentation of REDIRECT_REALLOC
* Fix negative heap size values reported in WARN
* Fix null pointer dereference in TRACE_TARGET
* Fix of GC_bytes_allocd increment in GC_generic_malloc_inner
* Fix overlapping region assertion in mark_some if malloc redirect on Linux
* Fix potential SIGSEGV on out-of-memory in gctest
* Fix signals delivery fail in find-leak mode if init from non-main thread
* Fix stack top/bottom print order in GC_push_all_stacks if stack grows up
* Fix store-and-dirty call in GC_CONS
* Fix unregistering of thread created by intercepted pthread_create on NaCl
* Fix unused GC_parse_version if no SOFT_VDB and no parallel mark on Linux
* Fix use of unset errno after pthread_create/sigmask calls
* Fix various typos in comments and documentation
* Increment allocated objects count after GC_GCJ_MALLOC() in gctest
* Invoke GC_oom_fn if GC_make_array_descriptor failed because of no memory
* Make Emscripten Asyncify feature optional
* Mention gctba library in README.cmake
* Prevent 'function should return a value' BCC error in CMake script
* Provide meaningful error message in case of Emscripten threaded build
* Reduce local variable scope in resend_lost_signals_retry for cppcheck
* Remove disable-warning options from WCC_MAKEFILE unrecognized by wcc v2.0
* Remove false warning of missing libpthread.so on Linux
* Remove redundant 'ifdef THREADS' around LOCK/UNLOCK in call_with_alloc_lock
* Remove redundant GC_ATTR_UNUSED for GC_mark_and_push
* Remove unused GC_old_sig_mask from gc_locks.h
* Replace WARN in GC_mark_some wrapper back to GC_COND_LOG_PRINTF
* Specify throw and noexcept for operator new/delete in gc_cpp.h for MS VC
* Support Hurd/x86_64
* Support client-defined stack pointer adjustment before thread stack push
* Suppress 'unreachable code' wcc warning in I_HOLD_LOCK assertion (CMake)
* Update autotools for release preparation (ac-2.71, am-1.16.5, m4-1.14.19)
* Update dump function name in GC_DUMP_REGULARLY variable documentation
* Use emscripten_stack_get_base instead of emscripten_scan_stack
* Use sbrk() to get OS memory on Emscripten
* Workaround 'writing into region of size 0' gcc warning in suspend_handler
* Workaround CSA null pointer dereference FP in invalidate_map of cord/de
* Workaround a malfunction of soft-dirty bits clearing on Power9
Files: