- npipe.c:
* Use RtlPrefixUnicodeString instead of RtlPrefixString with casts.
* Check results of RtlCreateUnicodeString and RtlPrefixUnicodeString, return FALSE if they fail and set an appropriate last error.
* Free the string created with RtlCreateUnicodeString instead of leaking memory.
* Fix a path type check (RtlPathTypeUncAbsolute instead of RtlPathTypeRootLocalDevice).
- path.c: I prefer seeing the default case at the end of the switch (no functional changes).
[NTOS:MM]
- Use RtlPrefixUnicodeString instead of RtlPrefixString with casts.
[RTL]
- RtlPrefixString acts on general PSTRINGs (which are not UNICODE).
- Remove extra spaces between names of functions and parentheses.
- Clarify the fact that we run over characters.
svn path=/trunk/; revision=60085
- Enable qmgr tests again. They all get skipped now instead of crashing. When npfs_new manages to run these, we know it's really working well ;)
CORE-6536
svn path=/trunk/; revision=60079
[NPFS-NEW]: Actually implement NpCancelWaiter instead of making it ASSERT.
[NPFS-NEW]: Critical fixes to NpAddWaiter and NpWaitForNamedPipe to fix logic flaws.
NPFS-NEW now behaves without any visible regressions, and exhibits only 2 failures in the kernel32 pipe winetest -- vs 120+ failures with the current NPFS driver. It has 0 ntdll pipe failures, and 0 kmtest pipe failures.
svn path=/trunk/; revision=60070
[NPFS-NEW]: Fix busted up NpGetNextRealDataQueueEntry and fix its usage.
These changes don't appear to "fix" any tests, but they were real bugs. Thanks to ThFabba.
svn path=/trunk/; revision=60068
- Zero out the Int10 Bios call arguments structure before calling Videoprt Int10CallBios
- Properly check for VESA BIOS return value from such calls
- Use VGA registers to update the palette when possible.
Now bochs goes up to 3rd stage when using this driver in conjunction with framebuf_new.
svn path=/trunk/; revision=60063
[CSRSRV]: CsrInsertThread should call this API to make sure the thread isn't dead already. Update the callers to handle this possible failure.
This behavior was obeserved in Windows' CSRSRV.DLL. It now works in ReactOS as well.
svn path=/trunk/; revision=60059
[NTOSKRNL]: Make MmCreateSection do the correct access checks and parameter checks, just like ARM3 does. For example, asking for PAGE_EXECUTE on a non-image file no longer requests SYNCHRONIZE and FILE_READ_DATA...it asks... FILE_EXECUTE. Cause you know... that's what the caller..wants. Anyway, this, among other things, fixes LdrValidateImageChecksum (and those annoying errors 0xC0000022 in the logs). SMSS is now checking the checksums of Known DLLs just like before (in Win7 they stopped doing this to improve boot performance, lol). It also makes Windows' SMSS happy.
svn path=/trunk/; revision=60055
Use _SEH2_YIELD when exiting from a _SEH2 block via a return or a goto (otherwise --> bug in the stack).
This limitation will be removed in SEH3, but it's not used at the moment.
svn path=/trunk/; revision=60052
Don't hardcode the first parameter of the bugchecks. Instead, I introduce a NpBugCheck macro and file identifiers
(their names conform to the existing name convention of the driver) so that, when bugchecking, we report as the
first parameter of the NPFS_FILE_SYSTEM bugcheck the file ID (in the high word) and the line (in the low word)
where the bugcheck was emitted.
See the MSDN article "Bug Check 0x25: NPFS_FILE_SYSTEM" - http://msdn.microsoft.com/en-us/library/windows/hardware/ff557436(v=vs.85).aspx
for more information.
Inspired from the existing ext2 driver.
Should be done for the other filesystems (FAT, NTFS, CDFS, Redirector, Mailslot, etc...).
svn path=/trunk/; revision=60051
