[Webkit-unassigned] [Bug 168126] [GTK] WebProcess should run cleanup on quit to release resources

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Mon Feb 13 09:41:56 PST 2017


https://bugs.webkit.org/show_bug.cgi?id=168126

--- Comment #7 from Zan Dobersek <zan at falconsigh.net> ---
(In reply to comment #0)
> When exiting MiniBrowser by closing the graphical window, WebPage::close()
> is called on the WebProcess, and it leads most of the time to a successful
> call of ~MediaPlayerPrivateGStreamerBase().

This should always be the preferred path. To what degree things can be uninitialized is to be seen I guess. We've had issues with graphics drivers and atexit handlers before.

> When exiting MiniBrowser from command line with Ctrl-C, SIGINT is sent to
> all process in the process group, and WebProcess quits without properly
> destroying its objects.
> When exiting MiniBrowser with killall (SIGTERM), the WebProcess also quits
> without destroying its objects.

This can get tricky. Ideally WebProcess would only be closed when the UIProcess signals it to do so. Sending SIGINT or SIGTERM directly at a WebProcess or NetworkProcess should also kill it.

But Ctrl-C is a bit special since it's super useful for development purposes, yet can't be worked around easily since it sends out SIGINT to all the processes in the process group, and AFAIU if we don't handle SIGINT in the WebProcess, SIGINT will kill that WebProcess before the UIProcess can shut it down when it handles its own SIGINT.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20170213/cfe90414/attachment-0001.html>


More information about the webkit-unassigned mailing list