[Koying]

Yes that's perfectly fine for small errors, the problem is what happens if you have too much data in your buffer or bigger errors, how can one delay the output. This is solved in v17 by the AddPause method for non IEC passthrough and by adding IEC silence packages in IEC passthrough. Your v16 implementation detects those, throws them out, sacrifices sync and tries to change the clock, right (*)? Therefore the reports in this forum about issues after seek are issues by design.

Well, the v17 code makes no difference between small and big errors, in this code. It's exactly the same clock tweakings that were deemed "hacky" in v16.
All the magic is to avoid errors, of course

Silence doesn't exist in a raw bitstream. Besides, our "silence" (= full zero frames) are just ignored by the IEC packager of Android (as it should be, as it's syncing on PT frames the same way we do). If you had access to logcat, you'd see a bunch of "x bytes ignored" if you pass them through.

I might again be wrong, but the v17 pauses smells like https://github.com/koying/xbmc/blob/addd...K.cpp#L530 , which is indeed required by AE, at least in v16.
I remind you that everybody here still has lollipop, and marshmallow solves a lot of issues, as you noticed. I have no lipsync errors, whether seeking or not.

Anyway, the point is not to compare the implementations, the point is still that it's presumptuous to say an api is shit because it doesn't fit Kodi's arch