[Webkit-unassigned] [Bug 183348] REGRESSION(r222772): [GTK][WPE] WebProcess from WebKitGtk+ 2.19.9x SIGSEVs in WebKit::WebProcess::ensureNetworkProcessConnection() at Source/WebKit/WebProcess/WebProcess.cpp:1127

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=183348

--- Comment #35 from Michael Catanzaro <mcatanzaro at igalia.com> ---
(In reply to Alex Christensen from comment #34)
> Comment on attachment 337489 [details]
> Patch
> 
> I think it's a bad idea to add code that's implicitly port-specific like
> this. Why are you so sure you want to keep r222772? 

My preference is to roll out both r222772 and r230481, and carefully reconsider what we're doing here. Having special process termination behavior for GTK/WPE will lead to more platform-specific problems in the future.

I've already pointed out why the motivation for r222772 and r230481 is incompatible with the WebKit multiprocess design. The goal there was to prevent leaks of hardware resources that are not released by process termination, but this simply cannot be right because the web process is *expected* to crash before releasing its resources.

> Why are these sync
> messages failing?  Could they be made more robust?

The sync messages are failing because the network process is already gone. I guess we could try to ensure that all web processes have terminated cleanly before attempting to bring down the network process, but I'm not sure if that would really be a good idea.

-- 
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/20180416/2b82c242/attachment-0001.html>