On Mon Sep 25 08:19:06 2023 +0000, Rémi Bernon wrote:
We used to have a "h264parse" hardcoded too, but it ended up being apparently unnecessary. Maybe there's a better way here, like for instance use find_element to optionally add a parser element?
Good point, it is indeed unnecessary; the MPEG splitter already has a mpegvideoparse (it's needed to calculate the AM_MEDIA_TYPE), no need for two. (Though deleting it introduces a chroma-site=(string)jpeg constraint on the output caps, causing the same format change issue as the colorimetry removal above.)
I expected the test suite to complain about the MPEG data being unparsed, but it doesn't. I guess avdec_mpeg2video has its own parser already. Will fix.