On 5/4/22 03:32, Rémi Bernon wrote:
diff --git a/dlls/winegstreamer/wg_transform.c b/dlls/winegstreamer/wg_transform.c index dca85e7366e..3af9822b016 100644 --- a/dlls/winegstreamer/wg_transform.c +++ b/dlls/winegstreamer/wg_transform.c @@ -114,6 +114,10 @@ static GstElement *transform_find_element(GstElementFactoryListType type, GstCap for (tmp = transforms; tmp != NULL && element == NULL; tmp = tmp->next) { name = gst_plugin_feature_get_name(GST_PLUGIN_FEATURE(tmp->data)); + /* Ignore mpg123audiodec for now, as it has issues with frame delay, + * e.g. it does not output any data until it receives 2 frames. */ + if (!strcmp(name, "mpg123audiodec")) + continue;
I'd really rather not do this. Does this end up being a problem in practice, and if so, how? (E.g. if all it does is break tests, I'd rather just make the tests less restrictive, or mark things as todo_wine.)
Actually, I find this surprising anyway. IIRC mpg123 acted completely synchronously when I wrote the mp3dmod tests, and just looking at the GStreamer source I don't see any reason why mpg123audiodec wouldn't work synchronously. (Maybe it's different for layer 3, but I can't imagine why...)
FWIW I intend to do a similar thing to disallow vaapi codecs for H264. They are prioritized over avdec_h264, when both are available, but they have issues and won't let us implement correct plane alignment easily.
It's a bit pointless to use them anyway as we don't support GPU buffers yet, and they could be re-enabled later when we do and for that case only, assuming the alignments are correct, or when they get fixed.
Disabling vaapi is a bit more reasonable, although it'd also be nice if we can find a way to work around vaapi's bugs that doesn't require blacklisting the filters entirely. Better yet if we can find a way to avoid using vaapi if and only if it's broken.