- Wine-Devel - list.winehq.org

Re: Wininet: InternetSetStatusCallback Fix
by Robert Shearman Sept. 13, 2004

Sept. 13, 2004

Robert Shearman wrote: > Hi, > > You can call InternetSetStatusCallback on a HTTP handle, not just a > InternetOpen handle, so we should search up through the parents to > find it. Oops, I just checked on MSDN and this patch is wrong. We should copy and store the status callback in every handle, not just in the top level one. I'll submit a new patch soon. Rob

1 0

PEB documentation
by Mike Hearn Sept. 13, 2004

Sept. 13, 2004

This is interesting information, but where did it come from? Is this structure described in some MS documentation? If so, does it have the full layout of the TEB as well? > Log message: > Robert Reif <reif(a)earthlink.net> > Document all the structure members up to SessionId in the PEB.

1 0

Re: winecfg (was Re: W->A calls)
by Mike Hearn Sept. 13, 2004

Sept. 13, 2004

> It appears that Mike has been doing a lot of work on this. > http://www.winehq.org/hypermail/wine-devel/2004/09/0221.html > So you probably first want to coordinate with him. Yes. I'm aware that I'm holding people up right now due to my non-connectedness. I'm actually living in a bed&breakfast with no net connection, so I only have web/email access at work. It's kind of odd - I've been doing some work on it lately, and suddenly people pop up who want to work on it after a long period of inactivity. Oh well :) > I would start by running "winecfg", and seeing what doesn't work, which > is a lot. For example, on the very first tab, you can select a "Windows > Version", but changing the setting doesn't actually change anything (it > should make/alter a registry entry). The winecfg code is in > wine/programs/winecfg. Fixed by me already. Actually, I've done quite a bit of work on it, I'd recommend waiting for 2 weeks at which point I'll have net access to my laptop again and can hopefully send off my work. At that point I'll be disappearing back into university life and doing CW work again so the coast will be clear for others to work on it. > A long discussion of things that are needed is in the thread starting: > http://www.winehq.org/hypermail/wine-devel/2004/08/0007.html I've sorted some of these already, especially in the apps tab. So ... if you guys can wait for 2 weeks I should (hopefully) have something to show soon. thanks -mike

1 0

RFC: ntdll critical section debug
by Robert Reif Sept. 13, 2004

Sept. 13, 2004

I have been running into deadlock problems in the direct sound dll and have been playing with the debug part of critical sections. I added a lock name to Spare[1] to give the critical sections names that will show up in the error messages just like wine's ntdll does. After looking at the ntdll code, it looks like this causes a leak because ntdll statically initializes it's locks and doesn't delete the debug part when it finds Spare[1] used. I also tried adding a linked list of critical section debug structs just like real windows does. Because ntdll statically initializes it's locks, they don't get added to the list. I manually stuff the critical section list lock into the list so it shows up and user locks show up but not the other internal ntdll locks. Is there a good reason to statically initialize ntdll locks or can they be initialized like normal locks? Here is a patch that adds the linked list of debug info: Index: dlls/ntdll/critsection.c =================================================================== RCS file: /home/wine/wine/dlls/ntdll/critsection.c,v retrieving revision 1.27 diff -u -r1.27 critsection.c --- dlls/ntdll/critsection.c 6 Sep 2004 20:26:23 -0000 1.27 +++ dlls/ntdll/critsection.c 10 Sep 2004 16:29:14 -0000 @@ -36,6 +36,51 @@ WINE_DEFAULT_DEBUG_CHANNEL(ntdll); WINE_DECLARE_DEBUG_CHANNEL(relay); +static LIST_ENTRY RtlCriticalSectionList; +static CRITICAL_SECTION RtlCriticalSectionLock; +static CRITICAL_SECTION_DEBUG RtlCriticalSectionDebug = +{ + 0, 0, &RtlCriticalSectionLock, + { &RtlCriticalSectionDebug.ProcessLocksList, + &RtlCriticalSectionDebug.ProcessLocksList }, + 0, 0, { 0, (DWORD)(__FILE__ ": RtlCriticalSectionLock") } +}; +static CRITICAL_SECTION RtlCriticalSectionLock = { &RtlCriticalSectionDebug, -1, 0, 0, 0, 0 }; + +void critsection_init() +{ + InitializeListHead( &RtlCriticalSectionList ); + InsertHeadList( &RtlCriticalSectionList, &RtlCriticalSectionDebug.ProcessLocksList ); +} + +static void CRIT_Dump() +{ + int i = 0; + PLIST_ENTRY mark, entry; + PCRITICAL_SECTION_DEBUG debug; + + RtlEnterCriticalSection( &RtlCriticalSectionLock ); + + mark = &RtlCriticalSectionList; + for (entry = mark->Flink; entry != mark; entry = entry->Flink) + { + debug = CONTAINING_RECORD(entry, CRITICAL_SECTION_DEBUG, ProcessLocksList); + + TRACE("RtlCriticalSectionDebug[%d]:\n", i++); + TRACE(" Type = %04x\n", debug->Type); + TRACE(" CreatorBackTraceIndex = %04x\n", debug->CreatorBackTraceIndex); + TRACE(" CriticalSection = %p\n", debug->CriticalSection); + TRACE(" ProcessLocksList.Flink = %p\n", debug->ProcessLocksList.Flink); + TRACE(" ProcessLocksList.Blink = %p\n", debug->ProcessLocksList.Blink); + TRACE(" EntryCount = %ld\n", debug->EntryCount); + TRACE(" ContentionCount = %ld\n", debug->ContentionCount); + TRACE(" Spare[0] = %08lx\n", debug->Spare[0]); + TRACE(" Spare[1] = %08lx (%s)\n", debug->Spare[1], debugstr_a((char *)debug->Spare[1])); + } + + RtlLeaveCriticalSection( &RtlCriticalSectionLock ); +} + inline static LONG interlocked_inc( PLONG dest ) { return interlocked_xchg_add( dest, 1 ) + 1; @@ -126,12 +171,16 @@ crit->DebugInfo->Type = 0; crit->DebugInfo->CreatorBackTraceIndex = 0; crit->DebugInfo->CriticalSection = crit; - crit->DebugInfo->ProcessLocksList.Blink = &(crit->DebugInfo->ProcessLocksList); - crit->DebugInfo->ProcessLocksList.Flink = &(crit->DebugInfo->ProcessLocksList); crit->DebugInfo->EntryCount = 0; crit->DebugInfo->ContentionCount = 0; crit->DebugInfo->Spare[0] = 0; crit->DebugInfo->Spare[1] = 0; + RtlEnterCriticalSection( &RtlCriticalSectionLock ); + InsertTailList( &RtlCriticalSectionList, &(crit->DebugInfo->ProcessLocksList) ); + RtlLeaveCriticalSection( &RtlCriticalSectionLock ); + + if (TRACE_ON(ntdll)) + CRIT_Dump(); } } crit->LockCount = -1; @@ -195,12 +244,19 @@ crit->LockSemaphore = 0; if (crit->DebugInfo) { + RtlEnterCriticalSection( &RtlCriticalSectionLock ); + RemoveEntryList( &(crit->DebugInfo->ProcessLocksList) ); + RtlLeaveCriticalSection( &RtlCriticalSectionLock ); + /* only free the ones we made in here */ if (!crit->DebugInfo->Spare[1]) { RtlFreeHeap( GetProcessHeap(), 0, crit->DebugInfo ); crit->DebugInfo = NULL; } + + if (TRACE_ON(ntdll)) + CRIT_Dump(); } return STATUS_SUCCESS; } Index: dlls/ntdll/ntdll_misc.h =================================================================== RCS file: /home/wine/wine/dlls/ntdll/ntdll_misc.h,v retrieving revision 1.48 diff -u -r1.48 ntdll_misc.h --- dlls/ntdll/ntdll_misc.h 15 Jul 2004 22:07:21 -0000 1.48 +++ dlls/ntdll/ntdll_misc.h 10 Sep 2004 16:29:14 -0000 @@ -50,6 +50,7 @@ extern void debug_init(void); extern void thread_init(void); extern void virtual_init(void); +extern void critsection_init(void); /* server support */ extern void server_init_process(void); Index: dlls/ntdll/thread.c =================================================================== RCS file: /home/wine/wine/dlls/ntdll/thread.c,v retrieving revision 1.21 diff -u -r1.21 thread.c --- dlls/ntdll/thread.c 23 Aug 2004 18:52:54 -0000 1.21 +++ dlls/ntdll/thread.c 10 Sep 2004 16:29:14 -0000 @@ -134,6 +134,7 @@ debug_info.str_pos = debug_info.strings; debug_info.out_pos = debug_info.output; + critsection_init(); debug_init(); virtual_init();

3 2

spying COM
by Saulius Krasuckas Sept. 13, 2004

Sept. 13, 2004

Am not sure would it be of any use for list members, but just some minutes ago I've discovered "The COM spy". Quite old stuff, but whatever: http://staff.develop.com/jasonw/comspy/default.htm

2 1

epoll patch - status?
by Shachar Shemesh Sept. 12, 2004

Sept. 12, 2004

Hi Alexandre, I'm sorry for being a pest about this. I was wondering whether there is anything that needs to be done on that one in order to facilitate committing this patch? Shachar -- Shachar Shemesh Lingnu Open Source Consulting ltd. http://www.lingnu.com/

5 14

Re: IShellFolder::CompareIDs partially undocumented take 2
by Vitaliy Margolen Sept. 12, 2004

Sept. 12, 2004

Is there are any problems with this patch? Tests do confirm it. Vitaliy Margolen Saturday, September 4, 2004, 9:48:41 PM, Vitaliy Margolen wrote: > Since my previous patch wasn't excepted here it is again. This time with more fixes > and tests. > Also it _is_ documented that CompareIDs returns result in lower 16 bits. And this function > returns only -1/0/1 as a result. > Vitaliy Margolen > changelog: > dlls/shell32/shlfolder.c > IShellFolder::CompareIDs - return only -1/0/1 in lower 16 bit.

1 0

W->A calls
by Joris Huizer Sept. 11, 2004

Sept. 11, 2004

Hello, By accident, I removed the emails of the thread I started, but I'll look it up if necessary; I found this old wine issue of some time ago.. I think the patch it has there could be helpfull http://winehq.org/?issue=229#About%20Converting%20W-%3EA%20Calls However, I don't know how to make simple tests.. as I'd really need to test wether the code I write works, and works properly :p How do I write tests for that? I'm running linux (in practice, I never boot into windows) so how should I do this (testing without windows environment)? How are tests supposed to run? I hope you can give some clues on that; I found there are many W->A calls in dlls/winmm/winmm.c (according to the janitorial page) so I might start there (hoping not all of them are so tricky ;) ) Thanks for suggestions on this, Joris

7 9

Re: Correct ConvertSidToStringSidW with test
by Dimitrie O. Paun Sept. 11, 2004

Sept. 11, 2004

On Sat, Sep 11, 2004 at 08:37:09AM -0700, Juan Lang wrote: > ConvertSidToStringSidW was incorrect for low-value > authorities; it put spaces in the resulting string. > This corrects that, and includes a test case for it. Content-Description: advapi32.diff Uniffied diff, please :) -- Dimi.

1 0

Re: PROPSHEET: Wizard resizing
by Filip Navara Sept. 11, 2004

Sept. 11, 2004

Filip Navara wrote: > I know this patch isn't really correct, but the old code isn't correct > either. Without it about 80% of the wizards were resized incorrectly > (or not at all) and the header line wasn't hidden sometimes. Even > though I would like to code the resizing correctly it will require > major code shuffling that I can't do now due to time constraints. So I > believe for now this patch is less evil than letting the code as is. > > (I can provide more details about how it should be implemented > correctly on request) > > - Filip > > Changelog: > First attempt at correcting wizard resizing. Scratch this patch please. I'll try to implement it correctly... Regards, Filip

1 0