Seek when nothing buffered fails to flush decoder and seek the demuxer |
|
Issue descriptionSee also comment #24 on internal b/33387059 for full repro case. Steps: 1. Append 10 seconds of audio, begin playing it 2. At about 5 seconds, abort(), remove(0, buffered end), set timestampOffset=0, set currentTime = 0 [[ Observe in debug logs that Demuxer doesn't see the seek to time 0, unexpectedly. ]] 3. Append different audio for time [0,10]. tell the element to play() (this second play() shouldn't be required for repro) [[ Observe a brief continuation of the originally appended, now removed audio (symptom of an unflushed-decoder, so ISTM the media renderer also never got the signal to seek to time 0 in step 2), followed by new media. Observe in debug logs that Demuxer is feeding media to the reader beginning where it left off earlier (not from time 0.) ]] So it looks like, for MSE, a seek at least when nothing is currently buffered, is suppressed from signalling the media pipeline (chunkdemuxer and renderers) when it shouldn't be.
,
Jan 19 2017
More detail: MSE extends HTMLMediaElement's calculation of HTMLMediaElement.seekable (see https://www.w3.org/TR/media-source/#htmlmediaelement-extensions), which is used in HTMLMediaElement seeking. |
|
►
Sign in to add a comment |
|
Comment 1 by wolenetz@chromium.org
, Jan 19 2017