- Wine-GitLab - list.winehq.org

[PATCH v3 0/1] MR10343: Draft: vbscript: Document expected parser conflicts with %expect 10.
by Francis De Brabandere (＠francisdb) March 25, 2026

March 25, 2026

The 10 shift/reduce conflicts are inherent ambiguities where bison's default shift gives the correct behavior: colon chains extend greedily, and Else/End If bind to the innermost inline If (dangling-else rule). -- v3: vbscript: Document expected parser conflicts with %expect 10. https://gitlab.winehq.org/wine/wine/-/merge_requests/10343

2 1

Re: [PATCH v2 0/4] MR10421: winegstreamer: Improve quartz MPEG-1 video seeking.
by Akihiro Sagawa (＠sgwaki) March 25, 2026

March 25, 2026

On Wed Mar 25 14:29:50 2026 +0000, Elizabeth Figura wrote: > Looks like there are a couple todo_wine in quartz:mpegaudio that need to > be removed. Thanks for pointing that out. I've sent a follow-up merge request, !10447. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10421#note_133731

1 0

[PATCH 0/1] MR10447: quartz/tests: Remove todo_wine from tests that succeed now.
by Akihiro Sagawa (＠sgwaki) March 25, 2026

March 25, 2026

Follow up 63d79461bff. Thanks to Elizabeth Figura for pointing this out. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10447

2 1

[PATCH v2 0/1] MR10446: vbscript: Fix crash in ReDim on uninitialized dynamic arrays.
by Francis De Brabandere (＠francisdb) March 25, 2026

March 25, 2026

Dim arr() followed by ReDim arr(N) crashed due to a NULL SAFEARRAY pointer dereference. -- v2: vbscript: Fix crash in ReDim on uninitialized dynamic arrays. https://gitlab.winehq.org/wine/wine/-/merge_requests/10446

2 1

[PATCH v35 0/1] MR10192: kernelbase: Handle extended length path prefix in GetLongPathNameW
by Vishnunithyasoundhar S (＠svishnunithyasoundhar) March 25, 2026

March 25, 2026

``` Support the \\?\ prefix by skipping it during path resolution while preserving it in the output. Added specific handling for extended UNC paths (\\?\UNC\) to mirror standard UNC behavior, ensuring these paths are treated as network shares and not local devices. Some paths that use the extended length path were filtered out as UNC paths - but not all of them are UNC. There can be UNC path that use extended length path but not all extended length paths are UNC paths. ``` Signed-off-by: Vishnunithyasoundhar S svishnunithyasoundhar(a)gmail.com -- v35: kernelbase: Handle extended length path prefix in GetLongPathNameW https://gitlab.winehq.org/wine/wine/-/merge_requests/10192

3 9

[PATCH 0/1] MR10445: tomcrypt: Treat FreeBSD like Linux
by Gerald Pfeifer (＠gerald) March 25, 2026

March 25, 2026

We are currently seeing ``` libs/tomcrypt/mpi.c:5785:24: error: implicit declaration of function ‘arc4random’ [-Wimplicit-function-declaration] 5785 | d |= ((mp_digit) arc4random()); | ^~~~~~~~~~ ``` even if including <stdlib.h> and setting various standard defines, so simply don't treat FreeBSD specially and go with the Linux code. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10445

2 1

[PATCH v3 0/2] MR10324: win32u: Fix deadlock when setting hdc_src to the window's self DC.
by Zhao Yi (＠Zhaoyi) March 25, 2026

March 25, 2026

Supersed !9180 and !9177 When hdc_src is the window's self DC, Within the NtUserUpdateLayeredWindow function, there exists the following call sequence: window_surface_lock()-\>NtGdiAlphaBlend()-\>window_surface_unlock(). However, the implementation of the NtGdiAlphaBlend() function calls the windrv_GetImage() function, which internally invokes lock_surface() and unlock_surface(). In this situation, window_surface_lock() is called first to acquire a lock. When lock_surface() is subsequently called to acquire the same lock, the initial lock has not yet been released, ultimately resulting in a deadlock. -- v3: win32u/tests: Add tests for hdc_src is window's self dc when calling UpdateLayeredWindow. https://gitlab.winehq.org/wine/wine/-/merge_requests/10324

4 10

[PATCH v5 0/2] MR10321: mfsrcsnk: Parse pixel aspect ratio from the avcC atom.
by Yuxuan Shui (＠yshui) March 25, 2026

March 25, 2026

More specifically the SPS bytes in the avcC atom. Note this only fixes the problem for winedmo but not gstreamer. This is because gstreamer 1) does not parse the SPS, 2) will set pixel-aspect-ratio in caps regardless if a pasp atom is present. So there is no way for us to distinguish between 1) pasp being present and having an aspect ratio of 1:1, and 2) pasp not being present but pixel-aspect-ratio having been set to 1:1 by gstreamer as fallback. And because this is the case, I made sure winedmo is tested separately. In winedmo's case, if pasp is not present, we can set the PixelAspectRatio.Denominator in MFVIDEOFORMAT to 0 to distinguish. -- v5: mfsrcsnk: Parse pixel aspect ratio from the avcC atom. https://gitlab.winehq.org/wine/wine/-/merge_requests/10321

3 7

[PATCH 0/1] MR10424: services: Cancel overlapped IO after timeout.
by Yuxuan Shui (＠yshui) March 25, 2026

March 25, 2026

Though WaitForSingleObject timed out, the IO operation is still running in the background. After the function process_send_command returns, the OVERLAPPED object referenced by this IO operation becomes invalid. Later, when this IO finally completes, server_select will use invalid stack memory. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10424

4 7

[PATCH v2 0/1] MR10437: ntoskrnl.exe/tests: Fix stack use-after-free of OVERLAPPED.
by Yuxuan Shui (＠yshui) March 25, 2026

March 25, 2026

The ioctls started in `test_cancel_thread` complete only after the function has already returned, thus the OVERLAPPED objects needed have become invalid. Make them global variables. This should be fine since we only start the next test thread after we have completed the previously started IO operations. -- v2: ntoskrnl.exe/tests: Fix stack use-after-free of OVERLAPPED. https://gitlab.winehq.org/wine/wine/-/merge_requests/10437

2 1