- Implement _RTC_GetErrDesc, _RTC_SetErrorFuncW, add stubs for _RTC_NumErrors, _RTC_SetErrorType, _RTC_SetErrorFunc
- Implement _RTC_DefaultErrorFuncW which will be used, when no other error func was set
- Implement _RTC_InitBase, which will be called from _RTC_Initialize, and which in turn calls either _CRT_RTC_INITW, if CRT was linked, or a local _CRT_RTC_INITW0, which in turn will return the error function to use. This allows us to use a custom error function for all modules that link to (MSV)CRT. Only user32 does not really fit into here, since it uses the same startup code but does not link to MSVCRT.
[CRT]
- Call _RTC_Initialize from __main before initializing the global constructors, but fall back to a dummy _RTC_NoInitialize(), when RunTmChk.lib was not linked. Now we properly initialize both our own as well as MS RunTmChk lib.
svn path=/trunk/; revision=64816
- Improve memory layout by moving the 32/64 bit stack to 0x7000-0xF000 and moving the BIOSCALLBUFFER up a bit. This gives us 56 KB additional space for freeldr itself. This allows to compile freeldr with /RTC1 (x86 only, the x64 version would get too big). And yes, it works.
- Implement FrLdrBugCheckWithMessage to get some useful messages on the screen and use it in MmCheckFreeldrImageFile
- Merge the .rtc section into the .text section, when we use RTC1
- Check the location of the Extended BIOS Data Area, add it to the memory descriptors and make sure we have enough space to put freeldr in. If the location of the EBDA is too low, print out a nice message on a blue screen :)
svn path=/trunk/; revision=64815
* Use the runtime checks globally in ReactOS now, not just for user mode modules.
* We no longer need the MS support lib, as Timo kindly implemented one for ReactOS.
* Rename the lib to runtmchk.
[PSEH][CRT]
* Remove some constrictions to seh.s source files, we no longer need them.
[KERNEL32]
* We no longer need to explicitly link to the runtime checks lib.
[FREELDR][KEYBOARD][NTOS][SETUPLDR]
* Link to the runtime checks lib when this mode is enabled.
CORE-8626
svn path=/trunk/; revision=64807
* Use set_module_type() instead of the combination of set_entrypoint(), set_subsystem() and set_image_base().
* Remove the now redundant bugcodes dependency.
[FRAMEBUF_NEW]
* Remove redundant bugcodes dependency.
svn path=/trunk/; revision=64806
* Let the minimum required CMake version be 2.8.12 so that *cough* people with old RosBE would be warned to update it! (I'm looking at you, Arch).
svn path=/trunk/; revision=64801
Bugfixing... Part 5/X:
- Actually, we don't care about the name in the file index. We don't need it to read the data attribute, it's unnamed. And the returned name was buggy (this still needs work though)
- When we are getting a file record which is a directory, there is NO data attribute. Don't attempt to read it. This prevents failure on directory entry return.
These fixes enable NTFS directory listing in ReactOS!
Even though, more fixes are required. So far, the metadata of the files are kind of broken. Explorer seems to be unable to list files.
This keeps anyway being a major step forward, as this was totally impossible a month ago ;-).
For these willing to see NTFS in action in ReactOS: http://www.heisspiter.net/~Pierre/rostests/NTFS_listing.png
svn path=/trunk/; revision=64800
Bugfixing... Part 4/X:
- Fix a nasty bug in NtfsLookupFileAt() (how did it work before?). The name parsing was wrong (no progress was being made) and thus was leading to an infinite loop in directory browsing.
- Fix a lovely bug coming from a non-documented feature in NTFS. To properly read the MFT index, you've to apply a mask. Do this to properly handles MFT record. This fixes returned MFT index which allows resuse.
- Do not allow returning MFT records < 0x10 for now. Not sure whether it should be allowed, but so far, these are MFT special records, so let's forget about it. IIRC, they are available on Windows. But trying to chase another bug for the moment.
This does not fix yet directory enumeration.
svn path=/trunk/; revision=64793
- Use a helper function for copying little chunks of memory (1, 2, 4 and 8 bytes) because a profiling of ntvdm showed that especially 1 and 2 bytes of memory were read the most, and calling RtlCopy/MoveMemory for intensively copying 1 or 2 bytes was shown to be inefficient. We also don't use directly intrinsics/builtins because the compiler cannot know in advance the size of the memory to be copied, it cannot perform the required optimizations. It was checked that using the builtin-memcpy or memmove of GCC when compiling the program in release+full optimization mode just embedded a call to _memcpy, and naively using the movsX intrinsics of MSVC does not do the job of "moving" memory taking into account for the possible overlaps. Therefore, for small sizes (<= 8 bytes), we use copy assignments, whereas for large sizes (and for 3, 5, 7, 9+ bytes) we use the regular method of calling RtlMoveMemory. We gain ~=10% speed with this optimization.
- Also I use >> and & for dividing by 4 and 2 instead of the regular / and % operations because they are not optimized otherwise by default by MSVC (they are however, if you explicitely enable optimizations).
svn path=/trunk/; revision=64792
Properly fix warnings on MSVC, introduced in r38302.
"- 'allowwarnings=true' -- not anymore", they said.
Advice by ThFabba.
svn path=/trunk/; revision=64790
- prevent a buffer underrun (fixes stray quotes being appended on autocomplete in some cases)
- check if the input string is not empty (fixes first file in directory getting skipped on autocomplete)
- fix some typos
CORE-8623
svn path=/trunk/; revision=64787
* Workaround for CORE-8525, CORE-6023, CORE-7735 and CORE-8635.
* Brought to you by V and Amine. Thanks to Timo and Jerome for the reviews.
CORE-8525 CORE-6023 CORE-7735 CORE-8635 #comment Please retest with r64777 or later.
svn path=/trunk/; revision=64777
Bugfixing... Part 3/X:
- Really release the allocated memory in NtfsFindFileAt() and NtfsLookupFileAt(). Don't attempt to release caller pointer!
This fixes crashes on directory enumeration failure. Even though this latter still doesn't work.
svn path=/trunk/; revision=64776
Bugfixing... Part 2/X:
- Please welcome yet another structure in our NTFS driver... The INDX buffer which is used for non resident index entries. Doc (incomplete) here: http://bos.asmhackers.net/docs/filesystems/ntfs/INDX.html
- Make use of it in NtfsFindMftRecord() to replace some totally obscure code.
- Add some asserts to verify we read somehow what we expect. One of the assert is not working properly... Hum! Need to find out why.
We don't go any farther so far. Good news though: a FIXME has been replaced by another FIXME.
NB: This would deserve being ported to FreeLDR to fix the same FIXME.
svn path=/trunk/; revision=64771
Handle int 0x2c (assertion failure) in KDBG by breaking into the debugger instead of ignoring it and pretending it was handled. Now NT_ASSERT works on GCC builds as well. (As good as things work in KDBG...)
svn path=/trunk/; revision=64770
