- Wine-GitLab - list.winehq.org

[PATCH v7 0/5] MR9843: wtsapi32: Fix session id and desktop handling when using CreateProcessAsUserW with tokens from WTSQueryUserToken
by Katharina Bogad (＠hackathi) Jan. 4, 2026

Jan. 4, 2026

This Merge request resolves [Wine-Bug #58588](https://bugs.winehq.org/show_bug.cgi?id=58588). The underlying problem was that services started on their own, non-interactive winstation, but within the current user's session. This is trivially not the case for system services on Windows: they run, by design, in other user contexts (e.g. NTAUTHORITY\SYSTEM). Whenever a service wants to start a process within another user's graphical session, it needs to set the session id to the correct value. This may be done using the WTS API. This MR implements this functionality in a (hopefully) wine-compatible way: 1. Patch wineserver so tokens can be transferred to other sessions 2. Implement NtSetInformationToken for TokenSessionId 3. Change the session of non-interactive services to session 0 (from the default session 1) in addition to setting their desktop to `__wineservice_winstation\Default` 4. Let CreateProcessAsUserW reset the desktop to the default value `__winsta0\Default` if the session of the to-be-created process is not a service session. This avoids implementing all user and session handling in-depth while still allowing the underlying functionality to work. A real-world user of this mechanism is the Dassault 3DEXPERIENCE launcher, which runs as a system service and wants to open interactive child processes within a user's scope. -- v7: kernelbase: Let CreateProcessInternalW overwrite the desktop if needed https://gitlab.winehq.org/wine/wine/-/merge_requests/9843

3 7

Re: [PATCH v29 0/1] MR9414: shlwapi: Handling of "C:\...xx" in PathCanonicalize[AW].
by Thomas Csovcsity (＠thc13) Jan. 4, 2026

Jan. 4, 2026

On Sun Jan 4 22:52:45 2026 +0000, Alexandre Julliard wrote: > Please don't rebase your MR every day, it's a waste of CI resources, and > makes it look like it's not ready. We are currently in code freeze so it > will have to wait anyway. ok. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9414#note_126394

1 0

[PATCH 0/1] MR9852: include: Add missing flags in D3D12_RENDER_PASS_FLAGS.
by Biswapriyo Nath (＠Biswa96) Jan. 4, 2026

Jan. 4, 2026

Required for https://github.com/PCSX2/pcsx2/commit/43e073a18dc480328c9bad59c85e6bfa9a45f… -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9852

2 1

[PATCH v6 0/5] MR9843: wtsapi32: Fix session id and desktop handling when using CreateProcessAsUserW with tokens from WTSQueryUserToken
by Katharina Bogad (＠hackathi) Jan. 4, 2026

Jan. 4, 2026

This Merge request resolves [Wine-Bug #58588](https://bugs.winehq.org/show_bug.cgi?id=58588). The underlying problem was that services started on their own, non-interactive winstation, but within the current user's session. This is trivially not the case for system services on Windows: they run, by design, in other user contexts (e.g. NTAUTHORITY\SYSTEM). Whenever a service wants to start a process within another user's graphical session, it needs to set the session id to the correct value. This may be done using the WTS API. This MR implements this functionality in a (hopefully) wine-compatible way: 1. Patch wineserver so tokens can be transferred to other sessions 2. Implement NtSetInformationToken for TokenSessionId 3. Change the session of non-interactive services to session 0 (from the default session 1) in addition to setting their desktop to `__wineservice_winstation\Default` 4. Let CreateProcessAsUserW reset the desktop to the default value `__winsta0\Default` if the session of the to-be-created process is not a service session. This avoids implementing all user and session handling in-depth while still allowing the underlying functionality to work. A real-world user of this mechanism is the Dassault 3DEXPERIENCE launcher, which runs as a system service and wants to open interactive child processes within a user's scope. -- v6: kernelbase: Let CreateProcessInternalW overwrite the desktop if needed services: start services with session id 0 wtsapi32: Make WTSQueryUserToken return a token with correct session id ntdll: Implement NtSetInformationToken for TokenSessionId https://gitlab.winehq.org/wine/wine/-/merge_requests/9843

2 5

Re: [PATCH v2 0/1] MR9847: winex11: Drop SCIM workaround. - approved
by Sven Baars (＠sbaars) Jan. 3, 2026

Jan. 3, 2026

This merge request was approved by Sven Baars. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9847

1 0

Re: [PATCH v2 0/1] MR9847: winex11: Drop SCIM workaround.
by Sven Baars (＠sbaars) Jan. 3, 2026

Jan. 3, 2026

On Fri Jan 2 22:31:23 2026 +0000, Byeongsik Jeon wrote: > Thanks for the comment! > I couldn't find a proper bugfix log in the SCIM git log, so I did some > additional testing. > In my previous tests, I had only tested with scim-hangul. I tested again > with scim-pinyin and SCIM built-in rawcode_imengine. In conclusion, the > issue seems to be a bug in the process_key_event of each imengine. > The `process_key_event` function of the problematic SCIM imengines are > implemented as follows: > ``` > if (key.is_key_release()) > return true; > ``` > In the non-problematic scim-hangul, it returns false. > By returning `false` in this situation, SCIM can call > IMdkit::IMForwardEvent and the X client can process the key release > event. This is the same for the ibus and fcitx imengines as well. > At the time of a4a5a2e, SCIM was the mainstream xim server, but today > SCIM is not being developed anymore, it has been replaced by ibus and > fcitx, so it doesn't seem necessary for Wine to continue this > workaround. It also creates additional issues for xim servers that work > correctly, as in #55849. > And, I didn't use git revert because of the changes in > handle_wm_protocols() since a4a5a2e. > v2: Add more commit message. Thanks for the extensive explanation. It looks reasonable to me to remove the workaround. It also sounds like the workaround is broken even when using SCIM with scim-hangul, which is even more reason to revert it. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9847#note_126379

1 0

[PATCH v2 0/1] MR9851: win32u: Return failure from ResizePalette called with a count of zero.
by Bernhard Übelacker Jan. 3, 2026

Jan. 3, 2026

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=59203 [Testbot run with this patch](https://testbot.winehq.org/JobDetails.pl?Key=161382) -- v2: win32u: Return failure from ResizePalette called with a count of zero. https://gitlab.winehq.org/wine/wine/-/merge_requests/9851

1 2

[PATCH 0/1] MR9851: win32u: Return failure from ResizePalette called with a count of zero.
by Bernhard Übelacker Jan. 3, 2026

Jan. 3, 2026

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=59203 -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9851

2 3

Re: [PATCH v2 0/1] MR9847: winex11: Drop SCIM workaround.
by Byeongsik Jeon (＠bsjeon) Jan. 2, 2026

Jan. 2, 2026

On Fri Jan 2 22:23:05 2026 +0000, Sven Baars wrote: > Thanks for looking into this. Did you also confirm that the workaround > was necessary for you with older versions of SCIM? When looking into > that bug I was unfortunately not able to do that. > Also, it would probably be better to use `git revert` to revert the > original commit that implemented the workaround. And then mention in the > commit message that this has been fixed in SCIM (and include the bug > link like you already did of course). > Thanks! Thanks for the comment! I couldn't find a proper bugfix log in the SCIM git log, so I did some additional testing. In my previous tests, I had only tested with scim-hangul. I tested again with scim-pinyin and SCIM built-in rawcode_imengine. In conclusion, the issue seems to be a bug in the process_key_event of each imengine. The `process_key_event` function of the problematic SCIM imengines are implemented as follows: ``` if (key.is_key_release()) return true; ``` In the non-problematic scim-hangul, it returns false. By returning `false` in this situation, SCIM can call IMdkit::IMForwardEvent and the X client can process the key release event. This is the same for the ibus and fcitx imengines as well. At the time of a4a5a2e, SCIM was the mainstream xim server, but today SCIM is not being developed anymore, it has been replaced by ibus and fcitx, so it doesn't seem necessary for Wine to continue this workaround. It also creates additional issues for xim servers that work correctly, as in #55849. And, I didn't use git revert because of the changes in handle_wm_protocols() since a4a5a2e. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9847#note_126369

1 0

Re: [PATCH v2 0/2] MR9850: d3d*/tests: Test how lighting is affected by world and view matrices. - approved
by Alexandre Julliard (＠julliard) Jan. 2, 2026

Jan. 2, 2026

This merge request was approved by Alexandre Julliard. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9850

1 0