On 4/27/22 12:50, Rémi Bernon wrote:
diff --git a/dlls/winegstreamer/h264_decoder.c b/dlls/winegstreamer/h264_decoder.c index 8bfa15529db..778729f576c 100644 --- a/dlls/winegstreamer/h264_decoder.c +++ b/dlls/winegstreamer/h264_decoder.c @@ -585,10 +585,20 @@ static const IMFTransformVtbl transform_vtbl =
HRESULT h264_decoder_create(REFIID riid, void **ret) {
- struct wg_format output_format = {.major_type = WG_MAJOR_TYPE_UNKNOWN};
- struct wg_format input_format = {.major_type = WG_MAJOR_TYPE_H264};
This feels a bit awkward. input_format is somewhat ill-formed [not that struct wg_format is highly formalized, but wg_format_to_caps_h264() at least expects the profile and level to be set], but output_format has a major type of UNKNOWN, presumably to avoid passing an ill-formed type.
@@ -210,9 +209,13 @@ NTSTATUS wg_transform_create(void *args) * based on the actual output caps now. Matching decoders with the * raw output media type should be enough. */
- media_type = gst_structure_get_name(gst_caps_get_structure(sink_caps, 0));
- if (!(raw_caps = gst_caps_new_empty_simple(media_type)))
- if (!(raw_caps = gst_caps_new_any())) goto out;
- if (gst_caps_get_size(sink_caps))
- {
const gchar *media_type = gst_structure_get_name(gst_caps_get_structure(sink_caps, 0));gst_caps_append_structure(raw_caps, gst_structure_new_empty(media_type));- }
And this will end up checking whether H.264 or WMA can decode to *anything*, which, while probably sufficient in practice, is less useful in theory than checking whether it can decode to, say, NV12 or PCM respectively.
I don't think that wg_format structs need to be fully specified (e.g. just enough to calculate the GstStructure name seems sufficient), although it'd be nice to avoid generating GStreamer FIXMEs under normal operation.