[Webkit-unassigned] [Bug 173261] Retry network process connection if we fail to re-establish it

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Mon Jun 12 14:24:06 PDT 2017


--- Comment #7 from Brady Eidson <beidson at apple.com> ---
(In reply to Antti Koivisto from comment #4)
> > Adding this somewhat arbitrary "wait 1/4 second, retry 4 times" introduces a
> > very mysterious behavior to consider and maintain going forward.
> I'm not sure what there is to maintain here.

The fuzzy behavior of "ah screw it, we'll just keep on trying" as opposed to maintaining software that works the first time.

> > Re-establishing the network process connection should succeed the first time.
> > If it doesn't, that's the bug we should fix.
> Any ideas?

Any ideas about what...? You filed this bug with the statement:
"WebProcess::networkConnection() sometimes fails and hits CRASH. It is not clear why it fails but the condition might be transient. It may make sense to retry a few times."

That is literally the only information I'm going off of here.

It stands to reason that WebProcess::networkConnection() is failing for a reason, yet I see no description of an exploration of that reason.

> > You mention the problem "might be transient"? Have you shown this patch to
> > actually produce results in practice?
> No. The best way to establish that is to land it and see if it helps with
> crash rates.

If the primary goal was to reduce this WebProcess crash, I would agree with this path of exploration.

But I think the primary goal is actually to find out why establishing the network process connection is failing, and this patch does nothing to explore that.

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/20170612/cedbe767/attachment.html>

More information about the webkit-unassigned mailing list