Bug 187171 – [GStreamer] mp4 video hosted on http server won't load

• Status: RESOLVED
• Resolution: WORKSFORME
• Component: Platform

[Philippe Normand]

https://sbc-qsystem-dev.pathfinder.gov.bc.ca/sbc_new.mp4

See also https://github.com/Igalia/meta-webkit/issues/18

[Philippe Normand]

This seems to be yet another issue with our internal GStreamer httpsrc element.

[Charlie Turner]

I had some reports that r228603 "[GStreamer] Push smaller buffers from HTTP source" was causing MP4 files hosted via HTTP to fail loading in a downstream installation. It seemed like it might be related to this. However, that doesn't seem the case here.

What I have noticed is that this test works just fine for a number of runs. However, when it finds its way in the web / network cache, an interesting thing starts to happen. I see,

0:00:01.403721513    21       0xaea790 DEBUG           webkitwebsrc WebKitWebSourceGStreamer.cpp:503:operator():<source> Started request
(NetworkProcess) retrieving https://sbc-qsystem-dev.pathfinder.gov.bc.ca/sbc_new.mp4 priority 2
(NetworkProcess) Retrieval: revalidation already in progress for 'https://sbc-qsystem-dev.pathfinder.gov.bc.ca/sbc_new.mp4':

and then, 60 seconds later,

(NetworkProcess) Speculative revalidation completed for 'https://sbc-qsystem-dev.pathfinder.gov.bc.ca/sbc_new.mp4':

and the file starts playing. If I rm -rf ~/.cache and rerun the test, it doesn't stall for a minute at the start until the resource ends up cached again. If I run with the file hosted on my machine, I don't see such pauses.

So it seems something to do with a combination of caching and "speculative revalidation", which I'm unfamiliar with at the moment.

Does this match the situation your end Philippe?

[Philippe Normand]

I wasn't the one affected by this issue. The original bug reporter is Gil, see the downstream Github issue mentioned in comment 0.

[Charlie Turner]

I can't reproduce this on trunk. https://trac.webkit.org/changeset/234497 may well have helped, but the issue described in the GitHub ticket I can not reproduce. I'm closing this, please reopen if you have more information running with a recent WebKit (r234497 or later would be best).