29133 2015-09-17 21:27:38 +0000 Reset Parser State algorithm, in 'sequence' appendMode, should trigger potential discontinuity adjustment logic 2015-10-13 15:33:31 +0000 1 1 1 Unclassified HTML WG Media Source Extensions unspecified PC Linux RESOLVED FIXED P2 normal --- 1 wolenetz wolenetz jdsmith mike pcotton public-html-media public-html-bugzilla oldest_to_newest 123249 0 wolenetz 2015-09-17 21:27:38 +0000 Currently, the following sequence of actions is allowed and produces strange results: SourceBuffer.appendMode = sequence; Append [0,10) Abort Append [100,110) // or Append [0,10) again Without the Abort ( == reset parser state), the expected result would be: [0,20) (assuming each append began with a random access point such that no frames were dropped). With the Abort and the current Reset Parser State algorithm that is appendMode-agnostic, the result would be: [0,10 with a large not-real-gap until 100,110) // or in the alternate case: [0,10) from the second Append [0,10). I believe the correct way to fix this problem is to include the following step in the Reset Parser State algorithm: 'If mode equals "sequence": Set group start timestamp equal to the group end timestamp.' I will prepare a PR shortly for review. 123250 1 wolenetz 2015-09-17 22:04:02 +0000 PR is pending review at https://github.com/w3c/media-source/pull/9 Jerry, please take a look. Thanks, Matt 123251 2 wolenetz 2015-09-17 22:42:44 +0000 fyi - Associated fix in Chromium is pending CR currently at https://codereview.chromium.org/1355673004/ 123267 3 jdsmith 2015-09-21 22:17:14 +0000 We see the same gap issue on sequence appends, and support making this change. I can approve the PR if there are no other objections. 123280 4 wolenetz 2015-09-22 19:02:34 +0000 Thank you, Jerry. Note, if the PR is merged, I'll subsequently share another PR to add a related test to MSE wpt as part of fixing this bug. 123680 5 pcotton 2015-10-13 15:33:31 +0000 Resolved via https://github.com/w3c/media-source/pull/9