[Webkit-unassigned] [Bug 196198] [GTK] Unexpected timeout in http/tests/media/video-play-stall-seek.html

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Thu May 27 11:08:36 PDT 2021


https://bugs.webkit.org/show_bug.cgi?id=196198

Enrique Ocaña <eocanha at igalia.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |eocanha at igalia.com

--- Comment #6 from Enrique Ocaña <eocanha at igalia.com> ---
I've ran this test in X11 Release and the test doesn't timeout (Lauro already mentioned this), but fails.

The test is making a request of a media file to the load-and-stall.cgi, which serves part of the file and then stalls on purpose. At some point a seek back to 0 is done and the test expects the readyState >= HAVE_CURRENT_DATA (meaning "part of the video is loaded, including position 0", which is true) and networkState == NETWORK_LOADING (the loading still hasn't finished, this isn't always true, as sometimes it's NETWORK_IDLE).

Debugged why the last condition isn't met and found that CachedResourceStreamingClient::loadFinished() is called (the download is declared to be finished forever because 0 bytes have been read). I went down through the layers down to the IPC that listens the NetworkProcess messages, continued in the NetworkProcess and found that g_input_stream_read_async() was returning 0 bytes because libsoup just behaves that way (or that's whay I could understand with my limited knowledge of libsoup). If no bytes are available immediately, then 0 bytes are returned and the upper layers just assume EOS.

With this in mind, I think the test should be changed and just avoid the networkState check. I don't think anybody would care much, since the test is skipped on win, glib, gtk-wayland and mac (nobody uses it?).

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20210527/b33ecda5/attachment.htm>


More information about the webkit-unassigned mailing list