May 2026 - Wine-GitLab - list.winehq.org

[PATCH 0/1] MR10929: gdi32: Validate count parameter in ExtCreateRegion
by Bartosz Kosiorek (＠gang65) May 17, 2026

May 17, 2026

-- https://gitlab.winehq.org/wine/wine/-/merge_requests/10929

2 1

[PATCH v2 0/2] MR10889: ole32: Validate cached clipboard window handle in get_clipbrd_window.
by Aaron Yourk (＠ayourk) May 17, 2026

May 17, 2026

The OLE clipboard caches a per-process window handle. When the STA thread that created this window terminates, the window is destroyed but the stale handle remains cached. Subsequent OleSetClipboard calls use this stale handle, causing OpenClipboard to fail with CLIPBRD_E_CANT_OPEN. Check whether the cached window is still valid using IsWindow() before returning it. If the window has been destroyed, clear the cache so a fresh window is created on the current thread. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=59519 -- v2: ole32: Validate cached clipboard window handle in get_clipbrd_window. ole32/tests: Add test for OLE clipboard reuse across STA threads. https://gitlab.winehq.org/wine/wine/-/merge_requests/10889

2 2

[PATCH v4 0/1] MR10906: sapi: Implement ISpVoice::GetStatus, SetPriority and GetPriority.
by Bohdan Tkachenko (＠BohdanTkachenko) May 17, 2026

May 17, 2026

Previously these three methods returned E_NOTIMPL. SetPriority/GetPriority store the priority on the voice and validate it against SPVPRI_NORMAL, SPVPRI_ALERT and SPVPRI_OVER. GetStatus populates SPVOICESTATUS from speech_voice state: - ulCurrentStream tracks the stream currently being processed by speak_proc (set at the start of the worker so it reflects the in-flight stream, and sticks at the last processed value when idle, per MSDN). Stored as last_stream_num on speech_voice. - ulLastStreamQueued reuses the existing cur_stream_num counter, which is incremented before each Speak queues a task. - hrLastResult is the HRESULT of the most recently completed speak (set at the end of speak_proc). - dwRunningState is derived from async_wait_queue_empty(&queue, 0) so it correctly reports SPRS_DONE both when idle and after a purge clears queued tasks. This unblocks Overwatch 2's Audio Captions accessibility TTS, which calls SetPriority during voice setup and tight-polls GetStatus for SPRS_DONE between Speak calls. Add tests covering priority round-trips, GetStatus with a NULL pointer, and status fields before, during and after sync and async Speak calls. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=59749 -- v4: sapi: Implement ISpVoice::GetStatus, SetPriority and GetPriority. https://gitlab.winehq.org/wine/wine/-/merge_requests/10906

2 1

[PATCH v6 0/4] MR10751: winscard: Implement SCardGetCardTypeProviderName and SCardListCards (smartcard database)
by Guillaume Raffin (＠TheElectronWill) May 16, 2026

May 16, 2026

These functions are necessary to make some smartcard-based programs work. They rely on the "SmartCard Database", which can be found in the windows registry, at \`HKLM\\SOFTWARE\\Microsoft\\Cryptography\\Calais\\SmartCards\`. I have added extensive unit tests and validated the implementation with a real application (an internal program based on Atos CardOS). As this is my first PR here, please don't hesitate to tell me if something's wrong! I was wondering if the database-related code should be moved to a dedicated file `winscard-database.c`? same for the tests?), let me know what you think. -- v6: winscard: Implement SCardGetCardTypeProviderNameA winscard: Implement SCardListCardsA winscard: Implement SCardListCardsW winscard: Implement SCardGetCardTypeProviderNameW https://gitlab.winehq.org/wine/wine/-/merge_requests/10751

2 9

[PATCH v4 0/3] MR10751: winscard: Implement SCardGetCardTypeProviderName and SCardListCards (smartcard database)
by Guillaume Raffin (＠TheElectronWill) May 16, 2026

May 16, 2026

These functions are necessary to make some smartcard-based programs work. They rely on the "SmartCard Database", which can be found in the windows registry, at \`HKLM\\SOFTWARE\\Microsoft\\Cryptography\\Calais\\SmartCards\`. I have added extensive unit tests and validated the implementation with a real application (an internal program based on Atos CardOS). As this is my first PR here, please don't hesitate to tell me if something's wrong! I was wondering if the database-related code should be moved to a dedicated file `winscard-database.c`? same for the tests?), let me know what you think. -- v4: winscard: Implement SCardGetCardTypeProviderNameA winscard: Implement SCardListCardsA winscard: Implement SCardGetCardTypeProviderNameW https://gitlab.winehq.org/wine/wine/-/merge_requests/10751

2 4

[PATCH v3 0/1] MR10923: oleaut32: Normalize the SLTG value of cParamsOpt to MSFT convention
by Wolfgang Hartl (＠wolfo.dev) May 16, 2026

May 16, 2026

Bug: https://bugs.winehq.org/show_bug.cgi?id=24387 The issue: The SLTG-format typelib in INV7.OCX of MS Money 2000 crashes on the call for the function CreateProperies() with a vararg VT_VARIANT|VT_ARRAY|VT_BYREF parameter that is omitted by the caller. The caller INV7.OCX CreateProperties() receives the wrong type and immediately crashes reading VT_ERROR instead a pointer to a SAFEARRAY. The root cause: The SLTG binary format stores cParamsOpt in a 6-bit field. SLTG uses the maximum value (0x3f / 63) as a sentinel with the meaning "vararg", like the MSFT format uses -1 for the same purpose. ITypeInfo_fnInvoke() checks cParamsOpt \< 0 at four places to detect vararg and handle the optional VT_VARIANT|VT_ARRAY parameter correctly to allocate an empty SafeArray placeholder. Since 63 \< 0 is false, this check never triggers for SLTG-compiled typelibs. The ITypeInfo_fnInvoke() function instead comes to the generic VT_BYREF placeholder path, which sets: VT_VARIANT|VT_BYREF {VT_ERROR: DISP_E_PARAMNOTFOUND} instead of the correct VT_VARIANT|VT_ARRAY|VT_BYREF {empty SafeArray}. The fix: Normalize the SLTG sentinel value to the MSFT convention at parse time in SLTG_DoFuncs: set cParamsOpt to -1 when it is 63. This makes the cParamsOpt \< 0 checks in ITypeInfo_fnInvoke work correctly for SLTG typelibs, and GetFuncDesc returns the Windows-documented canonical value (-1) for vararg to callers. Tests: Rerun MS Money 2000 with the lasted compiled version wine version 11.9 plus the fix in typelib.c. With the fix, DispCallFunc receives the correct argument VT_VARIANT|VT_ARRAY|VT_BYREF the reference to an empty SafeArray, CreateProperties() executes successfully and returns VT_DISPATCH. The function call in MS Money 2000 no longer crashes! I extensively searched the internet and did not find any documentation of this vararg defintion for SLTG format. But I sucessfully recreated typelib files in SLTG format in Windows with vararg and without vararg option for a test-function. I was able to produce them with an old midl version 3.01.75, later versions failed with internal compile error MIDL9008. `midl /win32 /oldtlb /tlb test3.tlb test_vararg_sltg3.odl ` `midl /win32 /oldtlb /tlb test3no.tlb test_novararg_sltg3.odl` The difference in the generated binary typelib files shows the relevant byte changed from 0xfe (6-bits 111111) with "vararg" to 0x80 (6-bits 000000) without "vararg". `diff <(xxd test3.tlb) <(xxd test3no.tlb)` `< 00000100: 0000 feff ffff 2c00 14fe 9900 1c00 3a00` `> 00000100: 0000 feff ffff 2c00 1480 9900 1c00 3a00` Then a test dump of both loaded tlbs showed in - Windows: cParamsOpt = -1 (vararg) and 0 (no vararg) - Linux/Wine: cParamsOpt = 63 (vararg, before the fix) and 0 (no vararg) -- v3: oleaut32: Normalize the SLTG value of cParamsOpt to MSFT convention https://gitlab.winehq.org/wine/wine/-/merge_requests/10923

2 1

[PATCH v2 0/1] MR10923: oleaut32: Normalize the SLTG value of cParamsOpt to MSFT convention
by Wolfgang Hartl (＠wolfo.dev) May 16, 2026

May 16, 2026

Bug: https://bugs.winehq.org/show_bug.cgi?id=24387 The issue: The SLTG-format typelib in INV7.OCX of MS Money 2000 crashes on the call for the function CreateProperies() with a vararg VT_VARIANT|VT_ARRAY|VT_BYREF parameter that is omitted by the caller. The caller INV7.OCX CreateProperties() receives the wrong type and immediately crashes reading VT_ERROR instead a pointer to a SAFEARRAY. The root cause: The SLTG binary format stores cParamsOpt in a 6-bit field. SLTG uses the maximum value (0x3f / 63) as a sentinel with the meaning "vararg", like the MSFT format uses -1 for the same purpose. ITypeInfo_fnInvoke() checks cParamsOpt \< 0 at four places to detect vararg and handle the optional VT_VARIANT|VT_ARRAY parameter correctly to allocate an empty SafeArray placeholder. Since 63 \< 0 is false, this check never triggers for SLTG-compiled typelibs. The ITypeInfo_fnInvoke() function instead comes to the generic VT_BYREF placeholder path, which sets: VT_VARIANT|VT_BYREF {VT_ERROR: DISP_E_PARAMNOTFOUND} instead of the correct VT_VARIANT|VT_ARRAY|VT_BYREF {empty SafeArray}. The fix: Normalize the SLTG sentinel value to the MSFT convention at parse time in SLTG_DoFuncs: set cParamsOpt to -1 when it is 63. This makes the cParamsOpt \< 0 checks in ITypeInfo_fnInvoke work correctly for SLTG typelibs, and GetFuncDesc returns the Windows-documented canonical value (-1) for vararg to callers. Tests: Rerun MS Money 2000 with the lasted compiled version wine version 11.9 plus the fix in typelib.c. With the fix, DispCallFunc receives the correct argument VT_VARIANT|VT_ARRAY|VT_BYREF the reference to an empty SafeArray, CreateProperties() executes successfully and returns VT_DISPATCH. The function call in MS Money 2000 no longer crashes! I extensively searched the internet and did not find any documentation of this vararg defintion for SLTG format. But I sucessfully recreated typelib files in SLTG format in Windows with vararg and without vararg option for a test-function. I was able to produce them with an old midl version 3.01.75, later versions failed with internal compile error MIDL9008. `midl /win32 /oldtlb /tlb test3.tlb test_vararg_sltg3.odl ` `midl /win32 /oldtlb /tlb test3no.tlb test_novararg_sltg3.odl` The difference in the generated binary typelib files shows the relevant byte changed from 0xfe (6-bits 111111) with "vararg" to 0x80 (6-bits 000000) without "vararg". `diff <(xxd test3.tlb) <(xxd test3no.tlb)` `< 00000100: 0000 feff ffff 2c00 14fe 9900 1c00 3a00` `> 00000100: 0000 feff ffff 2c00 1480 9900 1c00 3a00` Then a test dump of both loaded tlbs showed in - Windows: cParamsOpt = -1 (vararg) and 0 (no vararg) - Linux/Wine: cParamsOpt = 63 (vararg, before the fix) and 0 (no vararg) -- v2: oleaut32: Normalize the SLTG value of cParamsOpt to MSFT convention https://gitlab.winehq.org/wine/wine/-/merge_requests/10923

2 1

Re: [PATCH v4 0/1] MR10882: dsound: Replace device reference counting with a stop event. - closed
by Jon Koops (＠jonkoops) May 16, 2026

May 16, 2026

This merge request was closed by Jon Koops. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10882

1 0

Re: [PATCH v4 0/1] MR10882: dsound: Replace device reference counting with a stop event.
by Jon Koops (＠jonkoops) May 16, 2026

May 16, 2026

Superseded by !10919. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10882#note_140211

1 0

Re: [PATCH v4 0/1] MR10886: dsound: Remove unused DSOUND_renderers list and lock. - closed
by Jon Koops (＠jonkoops) May 16, 2026

May 16, 2026

This merge request was closed by Jon Koops. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10886

1 0