Bug 201726 - REGRESSION(r249205): [GStreamer][MSE] playbin3 assertion during YouTube front-page load: combiner_control_pad: assertion failed: (combine->sinkpad != NULL)
Summary: REGRESSION(r249205): [GStreamer][MSE] playbin3 assertion during YouTube front...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: WebKitGTK (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Nobody
URL:
Keywords:
: 201652 (view as bug list)
Depends on:
Blocks:
 
Reported: 2019-09-12 10:41 PDT by Charlie Turner
Modified: 2019-10-24 01:56 PDT (History)
4 users (show)

See Also:


Attachments
backtrace (73.21 KB, text/plain)
2019-09-12 10:41 PDT, Charlie Turner
no flags Details
GStreamer debug logging (153.39 KB, text/x-log)
2019-09-12 10:41 PDT, Charlie Turner
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Charlie Turner 2019-09-12 10:41:10 PDT
Created attachment 378656 [details]
backtrace

Backtrace available in attachment area. This is a racey failure unfortunately, but I see it quite often after the MSE refactoring.
Comment 1 Charlie Turner 2019-09-12 10:41:53 PDT
Created attachment 378657 [details]
GStreamer debug logging

Attach a run.log with lots of GStreamer logging
Comment 2 Philippe Normand 2019-09-15 02:06:21 PDT
*** Bug 201652 has been marked as a duplicate of this bug. ***
Comment 3 Michael Catanzaro 2019-09-17 07:29:05 PDT
IMO this is critical and we should revert the MSE rewrite if it can't be fixed ASAP.
Comment 4 Alicia Boya García 2019-09-17 08:52:59 PDT
(In reply to Michael Catanzaro from comment #3)
> IMO this is critical and we should revert the MSE rewrite if it can't be
> fixed ASAP.

I know. I'm working on this among other playbin3-related bugs. I don't mind if you revert the WebKitMediaSrc rework commit on releases, I already commented back in the day that it was too soon to release it in production.
Comment 5 Michael Catanzaro 2019-09-17 11:11:47 PDT
OK, I suggest we can leave it in trunk (it's OK if trunk is crashy for a couple weeks while you work on a fix), but revert for 2.26 branch (where it's not OK). Then you can have another six months to focus on making the rewrite nice and stable in trunk.
Comment 6 Michael Catanzaro 2019-09-21 08:20:06 PDT
(In reply to Michael Catanzaro from comment #5)
> OK, I suggest we can leave it in trunk (it's OK if trunk is crashy for a
> couple weeks while you work on a fix)

Well this might have been a bad suggestion; this crash is too much for Tech Preview too. Alicia, how inconvenient would it be for you if I revert this on trunk? Can you have it solved before 2.27.1 (three weeks)? I don't want this crashing in 2.27.1.
Comment 7 Michael Catanzaro 2019-09-21 14:24:56 PDT
Hm, I thought there was only one crash here, but trying to play a couple youtube.com videos 10x in a row just now, I got 10 crashes before I gave up. They were actually two different crashes, not this one. I've definitely seen this one many times in the past several days, but it's not what I'm hitting today. Reported bug #202078 and bug #202079.
Comment 8 Michael Catanzaro 2019-10-22 11:30:52 PDT
So I'm not sure why 2.27.2 was released before addressing these issues. The goal of Tech Preview is to be as new as possible while still avoiding known critical regressions, so I'll continue to hold Tech Preview to 2.26.
Comment 9 Carlos Garcia Campos 2019-10-23 00:11:00 PDT
(In reply to Michael Catanzaro from comment #8)
> So I'm not sure why 2.27.2 was released before addressing these issues. The
> goal of Tech Preview is to be as new as possible while still avoiding known
> critical regressions, so I'll continue to hold Tech Preview to 2.26.

The playbin3 support has been removed in 2.27.2. See bug #203078
Comment 10 Michael Catanzaro 2019-10-23 07:25:56 PDT
OK then, I'll try adding it to GNOME runtime.

So: can't this bug be closed now, or does Alicia want it open as a reminder to fix before relanding? What about bug #202078 and bug #202079?
Comment 11 Alicia Boya García 2019-10-24 01:56:57 PDT
It can be closed. The cause of the bug is playbin3 gets confused when several stream-collections are posted in quick succession; but for this and other reasons the next version of the WebKitMediaSrc patch will defer exposing any tracks until all SourceBuffer have received an initialization segment, emitting a single streams-collection message, which shouldn't cause this bug again.

I've replied to the other mentioned bugs too.