[Gc] Fix for MPROTECT_VDB
ivmai at mail.ru
Fri May 22 08:07:31 PDT 2009
The proposed patch refines the recent committed changes regarding MPROTECT_VDB.
* Makefile.direct: Update info for -DUSE_MUNMAP.
* Makefile.direct: Add info about new -DGC_PREFER_MPROTECT_VDB.
* doc/README.environment: Replace info about removed
GC_DISABLE_GETWRITEWATCH With info about new GC_USE_GETWRITEWATCH.
* doc/README.win64: Add info about mingw-w64 compiler.
* doc/README.win64: Add note about warnings suppression for VC++.
* include/private/gcconfig.h (GC_PREFER_MPROTECT_VDB): New macro
recognized (for the case when multiple VDB strategies available).
* os_dep.c (detect_GetWriteWatch): Remove "GC_DISABLE_GETWRITEWATCH"
environment var recognition.
* os_dep.c (detect_GetWriteWatch): Recognize "GC_USE_GETWRITEWATCH"
environment var and GC_PREFER_MPROTECT_VDB macro (only
* os_dep.c (GC_dirty_init): A check for defined(MSWIN32) is removed
for GC_gww_dirty_init() call (for Cygwin).
* win32_threads.c (GC_fault_handler_lock): Guard the declaration
with "if defined(MPROTECT_VDB) && !defined(MSWINCE)".
* win32_threads.c (GC_suspend): Ditto for GC_fault_handler_lock
* win32_threads.c (GC_push_stack_for): Add UNPROTECT(thread) before
setting last_stack_min value based on the hottest activation_frame.
* win32_threads.c (GC_get_next_stack): Add new local var 'thread'
and UNPROTECT(thread) (if 'thread' set) before updating
- Incremental collection (both MPROTECT and GWW) doesn't work on Cygwin yet;
- Incremental collection based on MPROTECT doesn't work with USE_MUNMAP and/or PARALLEL_MARK (Q: could this be improved?);
- Incremental collection based on MPROTECT (Win32) could crash an app if "Replaced other UnhandledExceptionFilter" warning is printed (or if the app replaces GC write_handler).
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 8778 bytes
Desc: not available
Url : http://napali.hpl.hp.com/pipermail/gc/attachments/20090522/7b75bf6a/diff83-0001.obj
More information about the Gc