September 2016 - Wine-Devel

Re: [PATCH 1/6] ddraw/tests: Add position tests for D3DPROCESSVERTICES_TRANSFORM.
by Marvin Sept. 3, 2016

Sept. 3, 2016

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=25546 Your paranoid android. === w1064 (32 bit ddraw1) === ddraw1.c:2240: Test failed: Expected (0,0)-(512,384), got (0,0)-(1024,768). === w1064 (64 bit ddraw1) === ddraw1.c:2240: Test failed: Expected (0,0)-(512,384), got (0,0)-(1024,768).

1 0

Where is my address space?
by Snorri Sturluson Sept. 3, 2016

Sept. 3, 2016

I’m trying to understand the memory usage of Wine. My focus is to make EVE Online run on the Mac as smoothly as possible, and one of the issues I’m running into is that the game runs out of memory in heavy scenes, especially when running at high resolutions on Macs with Retina screens. On Windows, a scene with 300 ships in space may be running in 2.1GB of page file usage, but the same scene under Wine will show a Virtual Memory size of 3.5GB or more. Once the Virtual Memory size (as shown by the Activity Monitor) gets too close to the 4GB mark I start getting errors from Wine, failing to allocate memory, usually leading to a crash. There is a significant difference even when running at the same resolution on both Windows and Mac, and further pronounced if I run in the full 5k resolution of an iMac. Can anyone explain this big difference in memory use between the Windows and Wine sessions? Or give me ideas on how I can figure it out? Or better yet, have suggestions on how to reduce it? Best regards, Snorri Sturluson CCP Games

4 9

"outsb" Emulation
by Morten Welinder Sept. 3, 2016

Sept. 3, 2016

I'm sorry, it's all my fault! Let me explain. A while ago -- well, 20+ years ago -- I was reading Wine code over and discovered a number of problems big and small. I have no recollection of what they actually were, but I shipped the list off to Alexandre who took at least some of my suggestions. Including one that was wrong. And that is why we have code like http://source.winehq.org/git/wine.git/blob/HEAD:/dlls/krnl386.exe16/instr.c… That FIXME is old -- older than the bug, actually. The answer is that no, it's not correct, but it used to be. "outsb" allows segment override, even if insb does not, see http://x86.renejeschke.de/html/file_module_x86_id_223.html The wine source file has gone though a few renames since then. I tracked it back to here: http://source.winehq.org/git/wine.git/history/58371aa361c7b119a5880b2f69d84… The first bad version occurs on 960114 -- so long ago that two-digit years were still a thing. I realize that it doesn't matter in practical terms. The code is unlikely to have been run multiple times over the past 20 years. Still, it's my fault. Morten Welinder PS: Please don't take my word for it. I really don't want to return in 2036 with an update. PPS: Keep up the good work!

1 0

Re: [2/2] imm32: Set the default IME window title and styles as in native.
by Alexandre Julliard Sept. 2, 2016

Sept. 2, 2016

Akihiro Sagawa <sagawa.aki(a)gmail.com> writes: > Signed-off-by: Akihiro Sagawa <sagawa.aki(a)gmail.com> > --- > dlls/imm32/imm.c | 6 ++++-- > dlls/imm32/tests/imm32.c | 9 +++++++++ > 2 files changed, 13 insertions(+), 2 deletions(-) It fails here: ../../../tools/runtest -q -P wine -T ../../.. -M imm32.dll -p imm32_test.exe.so imm32 && touch imm32.ok imm32.c:888: Test failed: got make: *** [imm32.ok] Error 1 -- Alexandre Julliard julliard(a)winehq.org

1 0

Implementing Fiber-Local Storage Callbacks: sharing data between DLLs
by John Sheu Sept. 2, 2016

Sept. 2, 2016

TL;WR: I'd like to share a struct definition between ntdll and kernel32, passed through the PEB. Is this kosher according to Wine style/design, and if so, which header should I stick it in? -- I'd like to have fiber-local storage callbacks in Wine. http://source.winehq.org/git/wine.git/commit/556fef3d http://source.winehq.org/git/wine.git/commit/ab11a5ae http://source.winehq.org/git/wine.git/commit/4e8f43f4 So, after writing the conformance tests, it's time to implement the feature :-) I've got the feature all done, except for one issue. The proximate issue is that when NtTerminateProcess is used on Windows, FLS callbacks are not called for threads that are aborted (and those threads do not get a DLL_THREAD_DETACH notification). However, if during DLL_PROCESS_DETACH the FLS slot is then freed, then the callbacks are called on behalf of those (already-terminated) threads by the process-detaching thread. The behavior appears to be thus: if the thread is exited cleanly (through LdrShutdownThread), then the FLS callbacks are called for that thread; otherwise on an unclean exit the FLS slots for the thread are transferred to some global list, and the callbacks for that list are called when FlsFree is later called. The underlying problem regards doing that transfer to the global list. LdrShutdownThread and its other friends reside in ntdll, so ntdll will have to know how to manage FLS. The FLS API however resides in kernel32, so kernel32 will have to know, as well. For this purpose I'd like to pass a common struct between ntdll and kernel32 through PEB.FlsListHead. What I see though is that include/winternl.h contains (almost all) publicly Microsoft-documented struct definitions, and I don't see anywhere else where such common structs are presently shared between two DLLs. So is it the case that the convention in Wine is to not do such things as sharing structs, or is this also done elsewhere and I just need to find the right header for it? Thanks, -John Sheu

4 6

Re: [3/3] imm32,user32: Create and destroy the default IME window implicitly. (try 9; resend)
by Sebastian Lackner Sept. 2, 2016

Sept. 2, 2016

On 23.08.2016 15:13, Akihiro Sagawa wrote: > > Try 9: > - Withdraw try 8 changes as we have ImmDisableIME now. > - Rebased and resolve conflicts against ImmDisableIME patch. > > Signed-off-by: Akihiro Sagawa <sagawa.aki(a)gmail.com> > --- > dlls/imm32/imm.c | 99 +++++++++++++++++++++++++++++++++++----------- > dlls/imm32/imm32.spec | 1 + > dlls/imm32/tests/imm32.c | 12 +++--- > dlls/user32/misc.c | 4 +- > dlls/user32/user_private.h | 1 + > dlls/user32/win.c | 26 ++++++++++++ > dlls/user32/win.h | 1 + > 7 files changed, 113 insertions(+), 31 deletions(-) This patch broke various user32 tests in virtual desktop mode, for example: $ (cd dlls/user32/tests/; make broadcast.ok) ../../../tools/runtest -q -P wine -T ../../.. -M user32.dll -p user32_test.exe.so broadcast && touch broadcast.ok broadcast.c:174: Test failed: Asynchronous message sent instead broadcast.c:152: Test failed: Returned: 0 broadcast.c:153: Test failed: Asynchronous message sent instead broadcast.c:174: Test failed: Asynchronous message sent instead broadcast.c:238: Test failed: Returned: 0 broadcast.c:239: Test failed: Asynchronous message sent instead broadcast.c:257: Test failed: Asynchronous message sent instead broadcast.c:238: Test failed: Returned: 0 broadcast.c:239: Test failed: Asynchronous message sent instead broadcast.c:257: Test failed: Asynchronous message sent instead broadcast.c:298: Test failed: Returned: 0 error 1460 broadcast.c:299: Test failed: Asynchronous message sent instead broadcast.c:309: Test failed: Returned: 0 error 1460 broadcast.c:310: Test failed: Asynchronous message sent instead broadcast.c:320: Test failed: Returned: 0 error 1460 broadcast.c:321: Test failed: Asynchronous message sent instead broadcast.c:332: Test failed: Asynchronous message sent instead make: *** [Makefile:197: broadcast.ok] Fehler 17 Could you please take a look?

2 4

Re: [PATCH 2/2] ole32/tests: Update tests to the new channel hook info structure.
by Huw Davies Sept. 2, 2016

Sept. 2, 2016

On 2 Sep 2016, at 11:32, Huw Davies <huw(a)codeweavers.com> wrote: > > Also add tests for method number and causality. Please ignore this, I’ve sent a better version that doesn’t trace() garbage when the new structs are in use. Huw.

1 0

Re: [PATCH] winejoystick.drv: Fix long time consuming task of looking for joysticks
by Sebastian Lackner Sept. 2, 2016

Sept. 2, 2016

On 01.09.2016 05:18, Bruno Jesus wrote: > + joysticks = scandir(p, &files, filter_js, sort_js); Usually malloc-like allocations should be avoided in the Wine source. I am not sure if Alexandre is willing to make an exception in this case because its only a very small amount of memory - nevertheless, replacing with a different directory enumeration mechanism should also not be too difficult. Regards, Sebastian

2 1

Re: user32/tests: Mark a test failure with todo_wine.
by Alexandre Julliard Sept. 2, 2016

Sept. 2, 2016

Sebastian Lackner <sebastian(a)fds-team.de> writes: > Signed-off-by: Sebastian Lackner <sebastian(a)fds-team.de> > --- > > I am getting very consistently test failures here: > > clipboard.c:199: Test failed: 0002: seqno unchanged > clipboard.c:199: Test failed: 0082: seqno unchanged > > Also, currently the clipboard tests trigger all kind of other race-conditions > and bugs in winex11. Are you already aware of those issues, Alexandre? Yes, the stricter tests reveal all sorts of issues with the X11 implementation. I'll get to that once the user32 side works correctly. I don't get the above failure though (not that it means much, sequence number support is also fairly broken). -- Alexandre Julliard julliard(a)winehq.org

1 0

Maximum OpenGL version
by Fabian Maurer Sept. 2, 2016

Sept. 2, 2016

Currently looking at the directx11 implementation, and I noticed I first need to set a "MaxVersionGL" in the registry to enable creation of 4.3 context and so DX11 support. What's the reason behind this? The change is from a1e718ccabc7c330dd16c7face78022965b03e20 (wined3d: Add a setting for the maximum OpenGL version to use). Why do we need to set the maximum version instead of just using the highest context we can get (and still use) ? Fabian Maurer

3 6