- Call tcp_shutdown() two different times for TX|RX shutdown to avoid the tcp_close() problem entirely
- Forcefully close the socket in LibTCPCloseCallback (not optimal, but good enough)
svn path=/trunk/; revision=58896
- Don't define mem_trim() to realloc() because mem_trim() must never move the buffer pointer
- Fix realloc() which was completely broken
- Properly handle a TX|RX shutdown
svn path=/trunk/; revision=58895
Implementation of pause, continue, and help commands for net service application. Patch by theflash.
Core-7131 #resolve
svn path=/trunk/; revision=58894
Unicodify and add localization support.
Patch by Lee Schroeder, simplified by me.
CORE-7142 #resolve #comment Committed in revision r58893, thanks!
[REGEDIT]
"Neutralize" common resources.
svn path=/trunk/; revision=58893
- Reformat the code for better readability. Sacrifice 80 chars max line width in favor of a readable code (sorry, but when the indentation eats 12 chars, formulas for calculating sizes are hard to fit into less than 70 chars, and they become totally unreadable).
svn path=/trunk/; revision=58890
- Start bringing sanity to FSRTL Large MCB implementation:
* Get rid of the GET_LIST_HEAD macro and define a proper private struct to access its members. That's much more flexible,and that's how it's usually done in hundreds of other places in the kernel. Otherwise we might just forget about structs and invent macros to access individual virtual fields in a byte array :)
* Going through the code (first, preliminary pass) and putting helper debug prints, especially when the code does not work as expected.
* Move variables definitions to the beginning of the function (thus get rid of up to local vars being redifined up to 3x times within a single function).
svn path=/trunk/; revision=58889
- Initialize SM_CXSCREEN and SM_CYSCREEN metrics in co_IntInitializeDesktopGraphics
- Also move some code from IntCreatePrimarySurface to co_IntInitializeDesktopGraphics
svn path=/trunk/; revision=58887
The idea then would be to have the following behaviour (when specifying the following options in the kernel command line):
/DEBUGPORT=COMi --> load KDCOM.DLL and use COMi port (i == 1,2,3,4) if possible.
/DEBUGPORT=FOO --> load KDFOO.DLL (useful for KDUSB.DLL, KD1394.DLL, KDBAZIS.DLL for VirtualKD, etc...)
/DEBUGPORT=ROSDBG:[COMi|SCREEN|FILE|GDB|...] --> load KDROSDBG.DLL which contains the ROS kernel debugger, and use COMi or SCREEN or... as output port.
svn path=/branches/kd++/; revision=58883
- turkish translation update by Erdem Ersoy: CORE-7025,CORE-7032,CORE-7033,CORE-7034,CORE-7071;
- italian translation update by Paolo Devoti: CORE-7122;
- use the same notepad caption title as in Windows 2003 for AHK test compatibility, patch by Edijs Kolesnikovics: CORE-7023;
- added polish translation of samsrv and wlanconf, updated polish translation of rapps, msgina, consrv, console and desk;
- convert msgina to compile resources for a defined language;
svn path=/trunk/; revision=58882
- Use SHELL_LINK_HEADER instead of defining a local structure. Patch by Dominik Hornung (dominikhornung at gmx dot de).
CORE-7099
svn path=/trunk/; revision=58881
- Add tests for ZwAllocateVirtualMemory, ZwCreateSection, ZwMapViewOfSection. Patch by Nikolay Borisov (nib9 at aber dot ac dot uk).
- Some tests are currently skipped because they crash or assert on ROS: See ROSTESTS-107, ROSTESTS-108, ROSTESTS-109, ROSTESTS-110.
ROSTESTS-105 #resolve
svn path=/trunk/; revision=58874
- Implement a mechanism for kmtests to retrieve information from user-mode. Patch by Nikolay Borisov (nib9 at aber dot ac dot uk).
- Currently supported function: QueryVirtualMemory
ROSTESTS-96 #resolve
svn path=/trunk/; revision=58873
- Rework FsRtlExpression test to be more maintainable by using a table. Patch by Víctor Martínez Calvo, a few comments by me.
CORE-7108 #resolve
svn path=/trunk/; revision=58864
* Account for the asm source files in all the HAL flavors. Should fix compiling them with VS (which regressed recently).
svn path=/trunk/; revision=58862
- Apply changes from http://code.reactos.org/cru/CR-30
- If an event object already exists, CreateEvent returns a valid handle to it. So that, if the returned handle is NULL, then automatically we are sure that it doesn't exist. Therefore remove unneeded checks.
svn path=/trunk/; revision=58856
- Handle VME for the PUSHF and POPF instruction in virtual mode
- There is no reason to set interrupt flag when handling POPF
- Fix a typo in POPF case : mask out the right variable
See http://www.rcollins.org/articles/vme1/ for reference
svn path=/trunk/; revision=58852
* Make sure that BaseAddress is at 64-k boundary in NtMapViewOfSection(). Fixes a couple NtMapViewOfSection tests.
* Brought to you by Zhan Jianyu (aka larmbr) with review and improvement by Timo.
CORE-7118 #resolve #comment A slightly improved fix was committed in r58851. Thanks ;)
svn path=/trunk/; revision=58851
- Apply similar fixes as in 58848 to RtlGetElementGenericTable (no need to make too fancy predecrements if they work incorrectly in this case when do/while loops were unnecessary at all).
svn path=/trunk/; revision=58850
- Do not set result variable NodeOrParent in RtlpFindGenericTableNodeOrParent in case the generic table is empty, just returning TableEmptyTree is enough.
- Fix improper enumeration of generic tables nodes. The way they were done previously clearly shows that noone was actually testing these APIs and a simple mistake (do/while instead of while loop) led to a NULL pointer access. Thanks to Pierre for developing MCB tests which revealed this problem.
Rephrasing Vladimir Lenin: "Test, test and again test!"
svn path=/trunk/; revision=58848
Reimplement (yes, once more...) all the DOS wildcards in FsRtlIs*InExpression().
This time with a better understanding of MSDN (doc can be sometimes quite cryptic...).
Which means that now the functions are passing all the tests and are even simpler.
svn path=/trunk/; revision=58847