November 2024 - wine-gitlab

[PATCH v2 0/2] MR6289: dsound: Reject NaN values in some functions.
by Aida Jonikienė (＠DodoGTA) 23 Nov '24

23 Nov '24

The tests aren't really tested though (they partially require interactive mode and they caused lag issues on one person's machine). -- v2: dsound/tests: Add NaN tests for floating-point 3D functions. dsound: Reject NaN values in floating-point 3D functions. https://gitlab.winehq.org/wine/wine/-/merge_requests/6289

2 2

[PATCH v3 0/2] MR6884: Draft: Iphlpapi.dll: Included SendARP
by Thi Garlet (＠garlett) 23 Nov '24

23 Nov '24

Returns MAC address by: - search local interfaces, or - send UDP packed that does not need linux privilege, then search local cache solve https://bugs.winehq.org/show_bug.cgi?id=9418 files updated - /dlls/iphlpapi/iphlpapi_main.c - /dlls/iphlpapi/tests/iphlpapi.c -- v3: Returns MAC address by: https://gitlab.winehq.org/wine/wine/-/merge_requests/6884

3 3

[PATCH 0/1] MR6890: win32u: Use boolean instead of bool as name of a union member.
by Gerald Pfeifer (＠gerald) 23 Nov '24

23 Nov '24

bool is a reserved identifier in C23 and would break with GCC 15 moving to -std=gnu23 (based on C23). Use boolean instead. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/6890

3 2

Re: [PATCH v5 0/1] MR5924: Draft: mfmediaengine: Implement SVR.
by Nikolay Sivov (＠nsivov) 22 Nov '24

22 Nov '24

Some things to test: * could we use sample allocator unconditionally? In other words, is it possible to get SVR used when no device manager was provided; * EVR has a circular buffer logic, so maybe we could copy from there; * it would be great if we could get rid of locking sample buffer for copies, and that should be possible if we always have d3d samples. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5924#note_88688

1 0

Re: [PATCH v5 0/1] MR5924: Draft: mfmediaengine: Implement SVR.
by Nikolay Sivov (＠nsivov) 22 Nov '24

22 Nov '24

Nikolay Sivov (@nsivov) commented about dlls/mfmediaengine/video_frame_sink.c: > + LONGLONG pts2; > + transfer_sample = TRUE; > + /* if the second sample we have is also OK, we'll drop the first and use the second */ > + sample_index_increment(&sample_read_index); > + if (sink->sample[sample_read_index] && sample_get_pts(sink->sample[sample_read_index], clocktime, &pts2) == S_OK) > + { > + *pts = pts2; > + IMFSample_Release(sink->sample[sink->sample_read_index]); > + sink->sample[sink->sample_read_index] = NULL; > + sink->sample_read_index = sample_read_index; > + } > + } > + else if (sink->presentation_sample && !sink->sample_presented) > + { > + hr = sample_get_pts(sink->presentation_sample, clocktime, pts); > + } I find this hard to understand. We'll need to document what's happening. For example, why does this check one more sample at all? Is that because buffer size is 2, or we always would check the adjacent one. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5924#note_88687

1 0

Re: [PATCH v5 0/1] MR5924: Draft: mfmediaengine: Implement SVR.
by Nikolay Sivov (＠nsivov) 22 Nov '24

22 Nov '24

Nikolay Sivov (@nsivov) commented about dlls/mfmediaengine/main.c: > { > media_engine_set_flag(engine, FLAGS_ENGINE_SHUT_DOWN, TRUE); > media_engine_clear_presentation(engine); > + /* critical section can not be held during shutdown, as shut down requires all pending > + * callbacks to complete, and some callbacks require this cs */ > + LeaveCriticalSection(&engine->cs); > IMFMediaSession_Shutdown(engine->session); > + EnterCriticalSection(&engine->cs); > } That's something else, unrelated to SVR work? -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5924#note_88686

1 0

Re: [PATCH v5 0/1] MR5924: Draft: mfmediaengine: Implement SVR.
by Nikolay Sivov (＠nsivov) 22 Nov '24

22 Nov '24

Nikolay Sivov (@nsivov) commented about dlls/mfmediaengine/main.c: > > IMFMediaEngineNotify_EventNotify(engine->callback, MF_MEDIA_ENGINE_EVENT_ENDED, 0, 0); > break; > + > + case MEEndOfPresentation: > + video_frame_sink_notify_end_of_presentation(engine->presentation.frame_sink); > + break; I think you can use MFSTREAMSINK_MARKER_ENDOFSEGMENT for this. The way you did it sort of bypasses normal session flow, there is potentially unknown number of pending samples in the pipeline once MEEndOfPresentation is received. This is only a problem if for some reason earlier node somehow requests more times than the sink. ENDOFSEGMENT is already used in EVR, and that seems to work. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5924#note_88685

1 0

Re: [PATCH v5 0/1] MR5924: Draft: mfmediaengine: Implement SVR.
by Nikolay Sivov (＠nsivov) 22 Nov '24

22 Nov '24

Nikolay Sivov (@nsivov) commented about dlls/mfmediaengine/main.c: > > static void free_media_engine(struct media_engine *engine) > { > + EnterCriticalSection(&engine->cs); This helper is called when refcount is already 0, nothing else should be using it at this point. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5924#note_88684

1 0

Re: [PATCH v5 0/1] MR5924: Draft: mfmediaengine: Implement SVR.
by Nikolay Sivov (＠nsivov) 22 Nov '24

22 Nov '24

Nikolay Sivov (@nsivov) commented about dlls/mfmediaengine/video_frame_sink.c: > + IMFMediaType *current_media_type; > + BOOL is_shut_down; > + IMFMediaEventQueue *event_queue; > + IMFMediaEventQueue *stream_event_queue; > + IMFPresentationClock *clock; > + IMFAsyncCallback *callback; > + float rate; > + enum sink_state state; > + unsigned int flags; > + IMFSample *sample[2]; > + IMFSample *presentation_sample; > + int sample_write_index; > + int sample_read_index; > + BOOL sample_request_pending; > + BOOL sample_presented; > + BOOL eos; These could be "bool", we use that in several modules already. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5924#note_88683

1 0

Re: [PATCH v5 0/1] MR5924: Draft: mfmediaengine: Implement SVR.
by Nikolay Sivov (＠nsivov) 22 Nov '24

22 Nov '24

Nikolay Sivov (@nsivov) commented about dlls/mfmediaengine/video_frame_sink.c: > + return CONTAINING_RECORD(iface, struct video_frame_sink, IMFMediaEventGenerator_iface); > +} > + > +static struct video_frame_sink *impl_from_IMFStreamSink(IMFStreamSink *iface) > +{ > + return CONTAINING_RECORD(iface, struct video_frame_sink, IMFStreamSink_iface); > +} > + > +static struct video_frame_sink *impl_from_IMFMediaTypeHandler(IMFMediaTypeHandler *iface) > +{ > + return CONTAINING_RECORD(iface, struct video_frame_sink, IMFMediaTypeHandler_iface); > +} > + > +static void video_frame_sink_samples_release(struct video_frame_sink *sink) > +{ > + for (int i = 0; i < ARRAYSIZE(sink->sample); i++) ARRAY_SIZE is more common throughout the codebase. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5924#note_88682

1 0

2025

2024

2023

2022

wine-gitlab November 2024