mkvalidator doesn't detect overlapping timestamps

[robUx4]

If you remux this file with mkvmerge, it will create BlockGroup of 1536ms, but the next BlockGroup will start at 256ms (rightfully).

This kind of incorrect behaviour should be detected by mkvalidator.

[mbunkus]

I don't think there's actually a bug here (or your description is wrong). I assume you're talking about track 1 as that is the only track for which mkvmerge will write something at time stamp 00:00:00.256.

Track 1 is an AC3 track. AC3 frames are always 1536 samples long. For a sampling frequency of 48000 this amounts to 32ms.

In the remuxed file the first BlockGroup for track 1 starts at time stamp 0 and contains eight frames. Therefore the BlockGroup's duration is 8 * 32ms = 256ms. The next BlockGroup for track 1 starts at time stamp 256ms – which is exactly correct.

I don't see a BlockGroup with a duration of 1536ms here.