* Deleting tasks & folders requires elevated privileges on Windows 7.
* Document SchRpcGetTaskInfo() calls that crash on Windows.
* Use a winetest context in the rpcapi loops.
* Use wide character string literals in rpcapi.c.
* Use __HRESULT_FROM_WIN32() to initialize constants.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/2516
--
v3: dinput: Rewrite IDirectInput8_EnumDevicesBySemantics.
dinput: Implement DIDBAM_PRESERVE BuildActionMap flag.
dinput: Check device type in BuildActionMap for specific semantics.
dinput: Load action map from registry before resetting guid.
dinput: Trace formats in (Build|Set)ActionMap and EnumDevicesBySemantics.
dinput/tests: Add more IDirectInput8_EnumDevicesBySemantics tests.
dinput/tests: Test BuildActionMap cases with multiple devices.
https://gitlab.winehq.org/wine/wine/-/merge_requests/2511
This series aims to improve GPU-side performance by avoiding VK_IMAGE_LAYOUT_GENERAL for textures that are used as render target and shader resource view. To do so, we have to transition them between VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL and VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL depending on use (and likewise for depth stencils).
It improves performance of Rocket League from 80 fps to 100 fps in a GPU limited configuration on my Radeon Polaris GPU.
This MR is marked as draft for now because I am not convinced by patch 3 yet. For actual submission I think I'll create separate MRs for patches 1-3 and 4-8.
Patch 2 introduces a validation layer error in the d3d11 tests that gets fixed in patch 3. No new test failures are introduced, although none of the existing ones are fixed either.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/2455
--
v2: dinput: Rewrite IDirectInput8_EnumDevicesBySemantics.
dinput: Implement DIDBAM_PRESERVE BuildActionMap flag.
dinput: Check device type in BuildActionMap for specific semantics.
dinput: Load action map from registry before resetting guid.
dinput: Trace formats in (Build|Set)ActionMap and EnumDevicesBySemantics.
dinput/tests: Add more IDirectInput8_EnumDevicesBySemantics tests.
dinput/tests: Test BuildActionMap cases with multiple devices.
https://gitlab.winehq.org/wine/wine/-/merge_requests/2511
Passing `NULL` or `INVALID_HANDLE_VALUE` into `IcmpCloseHandle` currently causes a segfault on wine, whereas it returns `FALSE` on windows.
It seems FFXIV occasionally does this.
--
v2: iphlpapi: Return FALSE for bad handles in IcmpCloseHandle
https://gitlab.winehq.org/wine/wine/-/merge_requests/2505
Windows has the LDAP_OPT_SSL option to turn SSL on or off after calling
ldap_init but before connecting to the server. OpenLDAP doesn't have
that option and instead expects a fully SSL connection to be requested
by passing one or more "ldaps://" URIs to ldap_init. However, OpenLDAP
also has an LDAP_OPT_URI option, which Windows does not have, that can
be used to change the URIs before connecting. And OpenLDAP already takes
care of converting "ldap://" or "ldaps://" to lowercase, so all we have
to do is find and replace that exact string in each URI in the list.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=54727
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/2508
This fixes two wine_todo-marked tests. However, I'm not entirely sure this is correct. The behavior of Report Count == 0 isn't entirely clear to me, but it does pass the test now. Some more tests might be needed.
cc @rbernon
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/2499