When calling webKitWebSrcStop in case of a seeking, the frame should not be reset (the source hasn't change). The frame may be used to get the network context. Some network stack (like Qt) are relaying on the network context.
Created attachment 84616 [details] Proposed patch.
This patch looks harmless :) I tested it on GTK, runned the media tests fine and did testing on Vimeo and Youtube and had no problem. The frame will now only be released on seek errors and in the PAUSED to READY transition, which sounds fine to me.
If the commit-queue lands this it might have issues with the ChangeLog as the patch doesn't seem to be based on HEAD, right?
(In reply to comment #3) > If the commit-queue lands this it might have issues with the ChangeLog as the patch doesn't seem to be based on HEAD, right? I don't think so. It has some magic stuff for the changelogs :)
CCing possible reviewers
Comment on attachment 84616 [details] Proposed patch. Clearing flags on attachment: 84616 Committed r80854: <http://trac.webkit.org/changeset/80854>
All reviewed patches have been landed. Closing bug.