Bug 208803 – REGRESSION(r257977): [GTK] http/tests/media/hls/video-cookie.html is a flaky failure

• Status: RESOLVED
• Resolution: FIXED
• Component: Media

[Diego Pino]

Diff:

--- /home/igalia/dpino/workspace/webkit-gtk/layout-test-results/retries/http/tests/media/hls/video-cookie-expected.txt
+++ /home/igalia/dpino/workspace/webkit-gtk/layout-test-results/retries/http/tests/media/hls/video-cookie-actual.txt
@@ -1,4 +1,4 @@

-EVENT(play)
+EVENT(stalled) TEST(false) FAIL
 END OF TEST

[Enrique Ocaña]

Thanks for finding this regression. It's caused by my patch from bug https://bugs.webkit.org/show_bug.cgi?id=205801 (doesn't happen without it, happens with it).

I'm having a look at how to fix it.

[Enrique Ocaña]

Created attachment 393049 [details]
Patch

[Enrique Ocaña]

Before the fix for bug 205801, when a streaming URL stalled a little bit or went slower than expected a duration for the stream was automatically (wrongly) set by WebKit GStreamer. Having a duration it was easier to trigger the end of the download when the actual end of a live stream was reached (and usually much before that).

After the fix for bug 205801 the live streams kept their infinite duration. When the server cut the stream, WebKit GStreamer didn't have a way to tell the difference between that situation and a real stall (because of network/server conditions), so it concluded it was a stall.

This patch fixed the test to "help" WebKit a little bit avoiding live streams. After all, testing for a cookie has nothing to do with live or non-liveness.

[Enrique Ocaña]

Created attachment 393052 [details]
Patch

[Enrique Ocaña]

Created attachment 393128 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 393128 [details]
Patch

Clearing flags on attachment: 393128

Committed r258205: <https://trac.webkit.org/changeset/258205>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/60271204>