[Webkit-unassigned] [Bug 174760] [GTK] Icon database error and crash

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=174760

--- Comment #10 from Carlos Garcia Campos <cgarcia at igalia.com> ---
(In reply to Michael Catanzaro from comment #8)
> Comment on attachment 316273 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=316273&action=review
> 
> I'm not sure I understand this fix. Is your goal to prevent the crash by
> ensuring that IconDatabase::performReleaseIconForPageURL is never called
> ever?

No, but when this is called no Image will be deleted.

> That is quite a hack, but I think that's fine as you say you're
> planning to delete all this code, 

It's a side effect, but the change is good in any case, it avoids creating an image for nothing when the page doesn't have a favicon.

> and I'm confident that will happen sooner
> rather than later.

Not sure how soon it will be, but definitely not in this release cycle.

> The IconDatabase has always been a mess anyway.

Indeed.

> I will test the change to see if it works... building now.

Thanks! I'll make a release tomorrow.

> > Source/WebKit/UIProcess/API/glib/IconDatabase.cpp:123
> >      // It's okay to delete the raw image here. Any existing clients using this icon will be
> >      // managing an image that was created with a copy of this raw image data.
> 
> Um... I'm not sure if this comment is really supposed to be in this function
> at all... but I would have added your new conditional above it rather than
> below it.

No, the data is removed when the image is destroyed, so setting it to nullptr would destroy any previous data.

-- 
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/20170724/7ff62732/attachment.html>