May 2024 - Wine-GitLab - list.winehq.org

[PATCH v2 0/3] MR5737: Draft: kernelbase: Use ERROR_DIRECTORY for FindFirstFileExW on invalid name.
by Brendan McGrath (＠redmcg) May 29, 2024

May 29, 2024

When `FindFirstFileA` is called with `<path>/<file>/*` (where file is expected to be a directory), Windows uses the `ERROR_DIRECTORY` error. This patch changes Wine's implementation to match Windows. This fixes a crash in Unity of Command II. -- v2: server: Don't return STATUS_OBJECT_NAME_INVALID on ENOTDIR. ntdll/tests: Test error code when NtOpenFile uses file as directory. https://gitlab.winehq.org/wine/wine/-/merge_requests/5737

3 4

[PATCH v3 0/3] MR5736: win32u: Prefer properties from Vulkan over fake ones.
by Krzysztof Bogacki (＠Saancreed) May 28, 2024

May 28, 2024

Currently, when GPUs are being added and display driver only provides a fake one (like e.g. winex11.drv does on Xwayland), its default name and empty UUID and PCI IDs will be preferred over ones reported by Vulkan. When Vulkan-only GPUs are added later, the resulting list will look like so: ``` Xrandr GPU 0x0000:0x0000 {00000000-0000-0000-0000-000000000000} NVIDIA GeForce RTX 4080 Laptop GPU 0x10de:0x27e0 {13938ad6-4925-b628-c068-30558bc4b489} ``` Instead, when default or empty properties were provided, we should prefer ones from Vulkan as they are more likely to be accurate: ``` Intel(R) Graphics (RPL-S) 0x8086:0xa788 {a7888086-0004-0000-0002-000000000000} NVIDIA GeForce RTX 4080 Laptop GPU 0x10de:0x27e0 {13938ad6-4925-b628-c068-30558bc4b489} ``` This is important for LUIDs to be later correctly assigned by winevulkan which searches for matching GPU by UUID. I'm not exactly sure if this implementation is the way to go (especially reporting and checking name for `"Wine GPU"`) but it's probably one of the simplest possible. Let me know if we already have some empty UUID around I could use for the comparison. -- v3: win32u: Prefer Vulkan UUIDs over empty ones. win32u: Prefer Vulkan PCI IDs over empty ones. win32u: Use common name for fake GPUs and prefer Vulkan name over it. https://gitlab.winehq.org/wine/wine/-/merge_requests/5736

3 4

[PATCH v5 0/4] MR5738: win32u: More DPI awareness context fixes.
by Rémi Bernon May 28, 2024

May 28, 2024

-- v5: win32u: Only keep DPI awareness context with window objects. win32u: Fix SetThreadDpiAwarenessContext. win32u: Pass the DPI awareness context in win_proc_params. win32u: Get the thread DPI context instead of the awareness. https://gitlab.winehq.org/wine/wine/-/merge_requests/5738

3 6

[PATCH 0/1] MR5704: makedep: Silence Clang warnings in bison-generated files.
by Jacek Caban (＠jacek) May 28, 2024

May 28, 2024

Depending on bison version and Clang config, generated files cause various warnings. In gitlab CI, there is an unused function in wbemprox. I don't see it on newer bison and LLVM, but there are other -Wunused-but-set-variable warning instead. This was partially fixed in upstream bison: https://git.savannah.gnu.org/cgit/bison.git/commit/?id=a166d5450e3f47587b98… The fix does not take into account MSVC targets, I tried upstreaming a fix for that, with no success: https://savannah.gnu.org/patch/?10420 https://lists.gnu.org/archive/html/bison-patches/2023-11/msg00000.html Together with !5628, this allows enabling `-Werror` in out CI. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5704

5 5

[PATCH v4 0/5] MR5744: winegstreamer: Fix various Cinepak issues
by Alfred Agrell (＠Alcaro) May 28, 2024

May 28, 2024

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=56698 I'm not hugely satisfied with some of those changes (especially how looping avi_decompressor_source_get_media_type now takes O(n^2) calls to ICDecompressQuery), but it gets the video in that issue working. If you've got a better idea for how to implement this stuff, do tell. -- v4: msvfw32/tests: Test that Cinepak rejects unsupported output types. iccvid: Reject unsupported output types. quartz/tests: Add Cinepak test to avi splitter. winegstreamer: Try DTS if PTS is absent. winegstreamer: Implement AM_MEDIA_TYPE to wg_format converter for Cinepak video. https://gitlab.winehq.org/wine/wine/-/merge_requests/5744

3 6

[PATCH 0/5] MR5745: wined3d: Remove miscellaneous unused definitions, part 3/4.
by Elizabeth Figura (＠zfigura) May 28, 2024

May 28, 2024

-- https://gitlab.winehq.org/wine/wine/-/merge_requests/5745

2 5

[PATCH v2 0/1] MR5720: wineps.drv: Only merge dmDefaultSource member of devmodes when a slot is found.
by Danyil Blyschak (＠blyss) May 28, 2024

May 28, 2024

There is a check earlier in the PSDRV_MergeDevmodes/merge_devmodes functions to update dmDefaultSource if a slot is found, and so this member should not be updated again with no such check. -- v2: wineps.drv: Only merge dmDefaultSource member of devmodes when a slot is found. https://gitlab.winehq.org/wine/wine/-/merge_requests/5720

4 3

[PATCH 0/1] MR5742: programs/conhost: Fix display of font preview (in 64bit mode).
by eric pouech (＠epo) May 28, 2024

May 28, 2024

Signed-off-by: Eric Pouech <epouech(a)codeweavers.com> -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5742

3 2

[PATCH v3 0/2] MR5734: mf: Implement audio capture device enumeration in MFEnumDeviceSources().
by Paul Gofman (＠gofman) May 28, 2024

May 28, 2024

-- v3: mf: Implement audio capture device enumeration in MFEnumDeviceSources(). https://gitlab.winehq.org/wine/wine/-/merge_requests/5734

4 5

[PATCH 0/1] MR5741: Draft: server: Implement key auto-repeat.
by Alexandros Frantzis (＠afrantzis) May 28, 2024

May 28, 2024

This is an early draft of an implementation of key auto-repeat in wineserver to get some feedback. Some open questions: 1. queue_keyboard_message requires a `current` thread, but we don't get one in timeout callbacks. At the moment I am manually setting `current` to the foreground thread, but I am wondering if that's acceptable or we should explore other ways forward (also see TODO in code). 2. This draft introduces a new server request to configure auto-repeat (`enable/delay/period`). I am thinking that for more straightforward integration with the keyboard repeat SPI parameters, the request should only support the `enable` flag and the server should query the SPI registry values to get `delay` and `period` when needed. I am wondering if there any caveats here since I don't see other code in the server querying registry values (well, except to implement the registry requests themselves). Also, I would hope that opening and caching the `HKCU\Control Panel\Keyboard` hkey would remove most of the cost of performing this operation (if that's even a concern at all). -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5741

4 9