New issue
Advanced search Search tips

Issue 889309 link

Starred by 1 user

Issue metadata

Status: Available
Owner: ----
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

Consider not injecting redundant or perhaps conflicting SPS/PPS info in AVC annex b conversion/injection

Project Member Reported by wolenetz@chromium.org, Sep 26

Issue description

A coded frame may already have SPS/PPS in-band. If it's a keyframe, at least Mp4StreamParser will currently inject SPS/PPS from the decoder configuration, regardless of any SPS/PPS which might already be in the frame.
 
Dan, is this purely a performance issue, or is it breaking functionality?
It's unknown if it's breaking things currently, but it was a source of problems for VTVDA until I worked around it. The workaround has been copied into DXVAVDA, although the vendor implementations could still have problems.
From chat with Dan, we may want to emit a full-on kConfigChange if we *detect* a new decoder configuration (SPS/PPS) that differs from the configuration from the init segment.

However, see also https://bugs.chromium.org/p/chromium/issues/detail?id=748250#c4 

Comment 4 by dbbrooks@chromium.org, Jan 18 (5 days ago)

Status: Available (was: Untriaged)
Assuming this is something that the team wants to do, in an effort to clear out the "Untriaged" bucket, marking as "Available".

@wolenetz, when you get a chance, could you confirm it has the correct pri level? (I'm helping out with the triage of Internals>Media issues, let me know if you have any feedback. Thanks.


Sign in to add a comment