- Keep APC delivery disabled while holding a Heap Lock, not just while acquiring/releasing it (Dedicated to Timo and Ged).
- Store the Heap Lock in non-paged pool, the only place where executive resources belong (Dedicated to patient people).
[RTL]
- Restructure/simplify/correct Heap and Heap Segment initialization (partially to cope with the changed Heap Lock interface).
- Restrict the location of Heap Segment headers to the base address of the Heap Segment (which frees up a whopping 60 bytes per Heap!).
- Cater for acquiring the Heap Lock exclusively or shared (the latter is only available in kernel-mode); only exclusive locking is used for now.
- Use a plain critical section to guard the Process Heap list, no reason to disguise it as a Heap Lock (saves us a handful more bytes).
svn path=/trunk/; revision=53761
Fix CsrProbeForRead and CsrProbeForWrite to actually access the memory, by using volatile. The compiler optimized the access away previously.
svn path=/trunk/; revision=53760
- Fix a number of MSVC/amd64 warnings/problems
- Remove code duplication in RtlCharToInteger and RtlUnicodeStringToInteger
- Fixup too large string length in RtlInitAnsiString, RtlInitUnicodeString
- Use symbolic constants instead of hardcoded values
- Fix overflow check in RtlInitUnicodeStringEx
svn path=/trunk/; revision=53746
- Merge changes from the themes branch. These changes will introduce test regressions for class functions and for TrackMouseEvent. The changes in the class functions are needed in order to let comctl32 register its own versions of built in controls. This is a hack and will be removed when we properly support class versions. The second change that causes test regressions is in TrackMouseEvent and it is needed in order to fix WM_MOUSELEAVE messages that were completely broken and not tested at all by wine's test suite.
svn path=/trunk/; revision=53745
- Make Vector parameter in HalEnableSystemInterrupt, HalDisableSystemInterrupt and HalBeginSystemInterrupt an ULONG, not an UCHAR
[NDK]
- 64bit fixes for HANDLE_TABLE, KPROCESS, SECTION_IMAGE_INFORMATION, MMADDRESS_LIST, MMVAD_FLAGS, MMVAD, MMVAD_LONG, MMVAD_SHORT, MEMORY_DESCRIPTOR, MEMORY_ALLOCATION_DESCRIPTOR, LdrVerifyMappedImageMatchesChecksum
- KDPC_DATA::DpcQueueDepth is signed on amd64, unsigned on x86
[NTOSKRNL]
- Fix hundreds of MSVC and amd64 warnings
- add a pragma message to FstubFixupEfiPartition, since it looks broken
- Move portable Ke constants from <arch>/cpu.c to krnlinit.c
- Fixed a bug in amd64 KiGeneralProtectionFaultHandler
svn path=/trunk/; revision=53734
Fix the overly complicated and buggy implementation of RtlIsDosDeviceName_Ustr based on the simple pre-r52687 implementation of RtlIsDosDeviceName_U. This fixes 14(!) wine tests.
svn path=/trunk/; revision=53726
RtlGenerate8dot3Name: Ignore spaces in long file names instead of converting them to underscores.
See issue #6385 for more details.
svn path=/trunk/; revision=53725
- Don't check for leading "." char in a filename, because it's valid in case of LFN. The driver is already quite broken, so it can't get more broken if this assert is commented out.
Fixes assertion in "kernel32_winetest module".
svn path=/trunk/; revision=53722
- Fix KiGeneralProtectionFault
- Don't bugcheck in KiUnexpectedInterrupt
- Remove debugprint in KiEndInterrupt, and instead assert, its called from the clock interrupt, since thats the only instance we handle right now
svn path=/trunk/; revision=53720
- add common amd64 asm trap macros
- fix amd64 version of HackEoi
- send EOI in amd64 trap exit code, so we avoid using HackEoi
- modify HalpClockInterruptHandler for better portability
- fix amd64 TscCalibrationISR
svn path=/trunk/; revision=53719
- get version.h and buildno.h out of precompiled headers, to avoid rebuilding the whole module each time we reconfigure the build
svn path=/trunk/; revision=53716
