August 2007 - Wine-devel - list.winehq.org

WINEDEBUG output issues
by Paul Vriens 15 Aug '07

15 Aug '07

Hi, I was trying to read a '+all,+relay,-syslevel,-gdi,-font' trace but several lines show something like: 000d:Cget_window_property000d:Call0x000d:, atom=000d:Call usePropertySheetInfo000d000000d:Call user32.CharNextA(0 {000d:Call user32.CharNextA(0040a04d "les\\Nmap") ret=0040551c The above is output from the same thread. I also have: 0009:trace:heap:RtlAllocateHeap (0x110000,00000002,0000014000d:Call user32.CharNe0009:trace:dc:GetDCState (0x474): retu000d:Ret user32.CharNextA() retval=004093c6 ret=0040551c Which mixes threads. Debug traces can already be hard to read but this is even more confusing. Any idea what the cause is (or better, the solution)? Cheers, Paul.

1 0

Re: [2/5] WineD3D: Optimize float constants in stateblocks
by Alexandre Julliard 14 Aug '07

14 Aug '07

Stefan Dösinger <stefan(a)codeweavers.com> writes: > From f49fd192ed2747663bdb0445dff9d602519fbc95 Mon Sep 17 00:00:00 2001 > From: Stefan Doesinger <stefan(a)codeweavers.com> > Date: Fri, 3 Aug 2007 20:26:29 +0200 > Subject: [PATCH] WineD3D: Optimize float constants in stateblocks > > --- > dlls/wined3d/device.c | 24 +++++++++++++- > dlls/wined3d/stateblock.c | 68 +++++++++++++++++++++------------------- > dlls/wined3d/wined3d_private.h | 4 ++ > 3 files changed, 62 insertions(+), 34 deletions(-) This one is broken: ../../../tools/runtest -q -P wine -M d3d8.dll -T ../../.. -p d3d8_test.exe.so device.c && touch device.ok wine: Unhandled page fault on write access to 0x000000f6 at address 0x61aad11c (thread 0039), starting debugger... WineDbg starting on pid 003a Unhandled exception: page fault on write access to 0x000000f6 in 32-bit code (0x61aad11c). Register dump: CS:0023 SS:002b DS:002b ES:002b FS:0063 GS:006b EIP:61aad11c ESP:0033fa90 EBP:0033fac8 EFLAGS:00010202( - 00 - -RI1) EAX:000000f6 EBX:61b3ae04 ECX:00000001 EDX:00000000 ESI:001bd4d0 EDI:001c6f78 Stack dump: 0x0033fa90: 001bd4d0 001bd4e4 00000001 00000000 0x0033faa0: 00000000 00000000 00000000 602d53b1 0x0033fab0: 00000000 00000000 0000002b 617da640 0x0033fac0: 617dad24 001bd4b8 0033faf8 617caebf 0x0033fad0: 00123d30 00000001 001bd4c0 001bd4b8 0x0033fae0: 606e97df 0000005c 617cae3b 60715808 Backtrace: =>1 0x61aad11c IWineD3DDeviceImpl_CreateStateBlock+0x3dc(iface=0x123d30, Type=0x1, ppStateBlock=0x1bd4c0, parent=0x1bd4b8) [/home/julliard/wine/wine/dlls/wined3d/device.c:478] in wined3d (0x0033fac8) 2 0x617caebf IDirect3DDevice8Impl_CreateStateBlock+0x8f(iface=0x121780, Type=0x1, pToken=0x33fe08) [/home/julliard/wine/wine/dlls/d3d8/device.c:1177] in d3d8 (0x0033faf8) 3 0x606e1260 func_device+0x2190() [/home/julliard/wine/wine/dlls/d3d8/tests/device.c:528] in d3d8_test (0x0033fe58) 4 0x606e8818 run_test+0x128(name=0x1103c5) [/home/julliard/wine/wine/dlls/d3d8/tests/../../../include/wine/test.h:389] in d3d8_test (0x0033fea8) 5 0x606e8ead main+0x14d(argc=<register ECX not in topmost frame>, argv=<register ECX not in topmost frame>) [/home/julliard/wine/wine/dlls/d3d8/tests/../../../include/wine/test.h:437] in d3d8_test (0x0033fed8) 6 0x606e958b __wine_spec_exe_entry+0x5b(peb=0x7ffdf000) [/home/julliard/wine/wine/dlls/winecrt0/exe_entry.c:36] in d3d8_test (0x0033ff08) 7 0x6041352e start_process+0xee(arg=0x0) [/home/julliard/wine/wine/dlls/kernel32/process.c:820] in kernel32 (0x0033ffe8) 8 0x60027997 wine_switch_to_stack+0x17() in libwine.so.1 (0x00000000) 0x61aad11c IWineD3DDeviceImpl_CreateStateBlock+0x3dc [/home/julliard/wine/wine/dlls/wined3d/device.c:478] in wined3d: movl %edx,0x0(%eax,%edx,4) 478 object->contained_vs_consts_f[j] = j; -- Alexandre Julliard julliard(a)winehq.org

2 1

Partial phase out of *winehq.com addresses
by Duane Clark 14 Aug '07

14 Aug '07

As mentioned about a month ago, we were planning to phase out the winehq.com email addresses, and use winehq.org exclusively. It looks like the few holdouts have converted over for awhile now, so I added a filter to the mailman program to silently drop anything mailed to winehq.com. If Jeremy wants to set the MTA (or whatever) to reject those addresses, that might be better. Then anyone accidently mailing to the wrong address would (presumably) get a delivery error message. And it would take a small portion of the load off mailman.

2 1

Re: Font problem + patch solution
by Abel Chiaro 14 Aug '07

14 Aug '07

Alexandre Julliard wrote: > penna(a)bb.com.br writes: > > > There is a patch proposed at the bug #7571 (http://bugs.winehq.org/show_bug.cgi?id=7571) > > It works very well and I did not experiment any regression. Maybe > > it is not the ideal patch and if it is not, maybe we can provide a fix for > > that. However, at least it fixes the font problem and did not introduce > > any undesirable efects. > > Is that possible to apply that patch at the main tree? Comments > > are also welcome! > > You should provide test cases that show that your patch is correct. I used Dmitry's tests [1] for this patch. [1] http://www.winehq.org/pipermail/wine-patches/2007-March/036706.html Shouldn't that be enough, I'll look into it again. - Pedro. -- "Caritas omnia vincit."

2 3

Re: WGL: GDI support is not advertised with double buffering
by Roderick Colenbrander 14 Aug '07

14 Aug '07

Wouldn't it be cleaner to cache the double buffering property in the WineGLPIxelFormat and set the gdi cap in there? That would skip another unneeded function. Roderick

1 1

Re: user32: LB_SETITEMDATA should return 1 on success, not 0
by Alexandre Julliard 14 Aug '07

14 Aug '07

Peter Verthez <peter.verthez(a)advalvas.be> writes: > diff --git a/include/winuser.h b/include/winuser.h > index 4297a8d..97363e0 100644 > --- a/include/winuser.h > +++ b/include/winuser.h > @@ -2428,6 +2428,7 @@ typedef struct tagSCROLLBARINFO > > /* Listbox message return values */ > #define LB_OKAY 0 > +#define LB_SETITEMDATA_OKAY 1 /* undoc. OK return for LB_SETITEMDATA */ Please don't invent constants that don't exist on Windows. In this case TRUE would work just as well. -- Alexandre Julliard julliard(a)winehq.org

2 1

Re: [d3d8/tests] Make tests run again on win95
by James Hawkins 14 Aug '07

14 Aug '07

On 8/14/07, Paul Vriens <paul.vriens.wine(a)gmail.com> wrote: > Hi, > > Again only to avoid a pop-up. Didn't fix/test the actual tests. > > Changelog > Make tests run again on win95 > Please change the subject of your emails to match the dll that you're patching. As it stands, it looks like you're patching advapi32 5 times, and we have to dig into the patch to see what dll you're actually changing. -- James Hawkins

2 2

Font problem + patch solution
by penna＠bb.com.br 14 Aug '07

14 Aug '07

Hi all, Since wine 0.9.31 there is a font problem (at least latin charset). There is a patch proposed at the bug #7571 (http://bugs.winehq.org/show_bug.cgi?id=7571) It works very well and I did not experiment any regression. Maybe it is not the ideal patch and if it is not, maybe we can provide a fix for that. However, at least it fixes the font problem and did not introduce any undesirable efects. Is that possible to apply that patch at the main tree? Comments are also welcome! Thank you in advance.

2 1

Problems with focus losing on virtual desktops
by Christian Authmann 14 Aug '07

14 Aug '07

Hi, after installing 0.9.42 I noticed some odd behavior when playing Diablo 2. I did a git bisect and nailed it down to this patch: http://source.winehq.org/git/wine.git/?a=commit;h=d836a5062141dd42293ed044d… winex11drv: Correctly react to focus loss away from Wine. I noticed the following changes: a) Using diablo 2's windowed mode: d2 will realize when it's losing focus again (as it did somewhere around 0.9.2x). As far as I can tell, this works reliable. (That's not a good thing for diablo 2, since d2 will minimize on losing focus and cannot be brought back to life afterwards. But that's a different bug.) >From now on, please read "Diablo 2 minimizes" as "Diablo 2 thinks it lost focus". b) Using d2 in fullscreen mode inside wine's virtual desktop: d2 will sometimes think it lost focus, even if it didn't. Moving the wine window around with hotkey+drag: d2 minimizes. Moving the window by the title bar: nothing happens. (using kde/kwin btw) Sometimes just clicking inside the window while playing will cause d2 to minimize, although I haven't found a reproducible pattern (i.e. do I have to press certain keys, do I have to double-click, etc). Activating a different window outside the virtual desktop? d2 stays open. Or opening another application via hotkey, or pulling down yakuake via hotkey. It stays open. But, if I reactivate the virtual desktop window afterwards: d2 minimizes. Ouch. c) virtual desktop again: when holding ALT to show the dropped items, then left clicking somewhere, d2 thinks that ALT has been released. This makes it quite hard to pick up items. This problem does not occur when using regular windowed mode or when changing the hotkey to something different - as far as I can tell only ALT in virtual desktop mode is affected. Note that I don't have any hotkeys on Alt (or Alt+click), I bound window dragging to the otherwised unused windows key. While the patch does as intended on regular windows, it's effects on virtual desktop windows can only be described as weird. I don't know how virtual desktops are supposed to work. At least for d2, it'd be useful if one application inside the virtual desktop always keeps focused (or rather, is told it'd still be focused). IMHO that's the purpose of virtual desktops: banning the apps inside a container where they shouldn't be influenced by outside stuff, including activating a different window. Not losing focus is also more sensible when virtual desktops are used as a way to ban fullscreen-apps into windows: most fullscreen applications don't expect to lose focus and behave weird. However, even if you think of virtual desktops simply as a means to group windows, it's currently broken: losing focus on reactivation is a little late. I have compiled the latest git version without this patch, and virtual desktop mode works fine again, so it's really down to this single line of code. I'm not far enough into the code to know any possible solution, but for immediate results a simple if (inside a virtual desktop) do old behaviour else do new behaviour should do the trick until a better solution is found. greetings, Christian

3 5

Re: kernel: add execution permission to DOS memory area
by Alexandre Julliard 14 Aug '07

14 Aug '07

Mikolaj Zalewski <mikolaj(a)zalewski.pl> writes: > Execution permission is needed because we store some (mostly > DPMI-related) code in UMB (in winedos/himem.c) that is executed for > each Win16 app. The user may put even more code there. Ntvdm.exe also > makes the DOS memory executable (except for some parts of the UMB) winedos is supposed to make that memory executable when needed. The executable protection was removed from kernel32 as it causes failures on platforms that don't allow write+exec. -- Alexandre Julliard julliard(a)winehq.org

1 0