January 2023 - Wine-gitlab - list.winehq.org

[PATCH 0/1] MR2067: wine32u: fix off-by-one bugs in MsgWaitForMultipleObjectsEx drivers
by Ivan Chikish (＠nekotekina) 30 Jan '23

30 Jan '23

When count = 0, returned value may become -1 which seems wrong in the case of successful process_events() call. Also MSDN suggests that `count` may be a valid return value in this case. When -1 is returned, it may cause SetLastError to -1 as well which is not a valid error code AFAIK. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=54405 -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2067

4 7

[PATCH v5 0/2] MR2051: include/msvcrt: Add noreturn attribute to _assert.
by Jinoh Kang (＠iamahuman) 30 Jan '23

30 Jan '23

This informs the compiler that no code will ever run after an assertion failure (unless NDEBUG is defined), which increases accuracy of compiler warnings and static analyses. -- v5: include/msvcrt: Add noreturn attribute to _assert. include: Add noreturn attribute to RtlRaiseStatus. https://gitlab.winehq.org/wine/wine/-/merge_requests/2051

3 4

[PATCH 0/1] MR2066: gdi32/tests: Mark another D3DKMTCheckOcclusion() test as flaky.
by Hans Leidekker (＠hans) 30 Jan '23

30 Jan '23

It shows the same behavior as the one marked by commit 2d19baa6b576. Wine-Bug: https://bugs.winehq.org//show_bug.cgi?id=51313 -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2066

3 2

[PATCH 0/2] MR2065: ws2_32/tests: Fix some test failures on Windows 11.
by Hans Leidekker (＠hans) 30 Jan '23

30 Jan '23

-- https://gitlab.winehq.org/wine/wine/-/merge_requests/2065

2 2

[PATCH 0/7] MR2064: Implement kernelbase.EnumProcessModulesEx
by eric pouech (＠epo) 30 Jan '23

30 Jan '23

This series implements kernelbase.EnumProcessModulesEx(). This series: - implements EnumProcessModulesEx() - ensure it passes a bunch of new tests - fix GetModuleBaseName, GetModuleFileNameEx, GetModuleInformation() to cope with HMODULES returned from EnumProcessModulesEx(). Notes: as EnumProcessModules() doesn't return the same list as Windows on a Wow64 setup, and that there's no use of EnumProcessModulesEx() in Wine code, this first series will implement EnumProcessModulesEx() and if accepted, it will be followed by a second serie which, firstly, adapt all places in Wine code with bad expectations on EnumProcessModules(), and secondly reimlement EnumProcessModules() using EnumProcessModulesEx(). -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2064

2 7

[PATCH v2 0/1] MR2063: dinput/tests: Dynamically allocate async operation handlers.
by Rémi Bernon 30 Jan '23

30 Jan '23

Instead of passing them from the stack. The async operation is executed in a thread pool, which might Release its reference asynchronously and call a stale function pointer if it is sequenced after the stack invalidation. -- v2: dinput/tests: Dynamically allocate async operation handlers. https://gitlab.winehq.org/wine/wine/-/merge_requests/2063

1 1

[PATCH v2 0/2] MR2055: win32u: Make more definitions hidden.
by Zebediah Figura (＠zfigura) 30 Jan '23

30 Jan '23

-- v2: win32u: Make set_visible_region() hidden. win32u: Make send_message_timeout() hidden. https://gitlab.winehq.org/wine/wine/-/merge_requests/2055

3 3

[PATCH 0/1] MR2059: combase: Generate apartment OXID in rpcss
by Aaron Hill (＠Aaron1011) 30 Jan '23

30 Jan '23

Currently, apartment OXIDs are generated using the process ID (and the thread ID for a single threaded-apartment). This means that if an apartment is destroyed and re-created (using `CoUninitialize` followed by `CoInitializeEx`), the newly created apartment will end up with the same OXID as the old apartment. However, testing shows that Windows will generate a *new* OXID when an apartment is created in the above manner. This patch uses `rpcss` to generate a new OXID. The current process id is combined with an incrementing counter stored in rpcss. This ensures that re-creating an apartment will not re-use an OXID. Additionally, this fixes an issue that caused the .NET 4.8 installer to become stuck during the downloading stage under Wine. The installer appears to perform the following actions: 1. Call `IBackgroundCopyJob_SetNotify` interface on a BITS job. This causes us to create a proxy (in the other process hosting 'qmgr') for the `IBackgroundCopyCallback` pointer passed as a parameter. 2. Trigger MTA apartment re-creation (in the process running the setup, *not* the process with the `IBackgroundCopyCallback` proxy) through `CoUninitialize` followed by `CoInitializeEx`. 3. Call `IBackgroundCopyJob_SetNotify` on a newly created job, but with the same `IBackgroundCopyCallback` pointer parameter. When we deserialize the pointer passed to `IBackgroundCopyJob_SetNotify`, we will end up re-using the same `proxy_manager` that we created for the previous `IBackgroundCopyCallback`. This is due to the fact that the OIDs happen to match (due to the fact that the .NET 4.8 setup appears to perform actions in the same order between the old and new apartments), and the apartment OXIDs match as explained above. above. As a result, we will use the old IPID when we send RPC packets using this `proxy_manager`. However, the new and old IPIDs will *never* match, since their generation process includes `RtlGenRandom`. This will cause a fault packet to be generated on the listening side of the RPC connection. By avoiding re-using OXIDs across re-created apartments, we ensure that the proxy side will never incorrectly re-use a stale `proxy_manager`. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2059

3 4

[PATCH 0/2] MR76: vkd3d-shader/hlsl: Support rsqrt() and distance() intrinsics.
by Nikolay Sivov (＠nsivov) 30 Jan '23

30 Jan '23

-- https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/76

3 7

[PATCH 0/1] MR2057: winepulse: Make mult_alaw_sample() and mult_ulaw_sample() hidden.
by Zebediah Figura (＠zfigura) 30 Jan '23

30 Jan '23

-- https://gitlab.winehq.org/wine/wine/-/merge_requests/2057

3 2