These patches finish moving targa file handling into d3dx9.
--
v3: d3dx9: Add support for decoding targa files with a color map in d3dx9.
d3dx9: Add support for decoding targa files with run length encoding in d3dx9.
d3dx9: Add support for decoding targa files with different pixel orders in d3dx9.
d3dx9: Add support for loading basic targa images without WIC.
d3dx9/tests: Add more tests for loading targa files.
https://gitlab.winehq.org/wine/wine/-/merge_requests/6688
On Fri Oct 18 14:05:23 2024 +0000, Rémi Bernon wrote:
> I see, I think this should not be necessary anymore since !6573, the new
> display mode will be written to the registry and other processes/threads
> should get their cache invalidate immediately.
> Later on, when the XRandR events are processed the cache might get
> invalidated again but should end up with the same display settings.
The issue was specifically with seeing mode changes from other processes. The test calls `EnumDisplaySettings()` in a loop waiting for the expected mode change. If we don't process `RRNotify` events in `xrandr14_get_id()` we won't see the new mode until some thread processes the messages. Since there are no such threads in the test process, the test will fail.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5060#note_85467