Bug 165280 – Rare crash when preparing to print

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit Misc.

[Brent Fulgham]

Crash trace statistics indicate a certain amount of nullptr dereferencing during some print operations. Inspecting the code revealed a number of places where the FrameView pointer was accessed without null checking, even though we null check this value everywhere else in the Frame class.

This patch adds additional checks to this code to try to reduce the cases where users encounter this problem.

[Brent Fulgham]

Created attachment 295899 [details]
Patch

[Brent Fulgham]

<rdar://problem/28240254>

[Brent Fulgham]

Created attachment 295900 [details]
Patch

[Tim Horton]

Comment on attachment 295900 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=295900&action=review

> Source/WebCore/page/Frame.cpp:899
> +            view->setCanHaveScrollbars(owner->scrollingMode() != ScrollbarAlwaysOff);

Seems a bit odd, since we just created and set the view. And still have a reference to it.

[Brent Fulgham]

(In reply to comment #4)
> Comment on attachment 295900 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=295900&action=review
> 
> > Source/WebCore/page/Frame.cpp:899
> > +            view->setCanHaveScrollbars(owner->scrollingMode() != ScrollbarAlwaysOff);
> 
> Seems a bit odd, since we just created and set the view. And still have a
> reference to it.

Oh! You're right. I was just blindly looking for view() calls. I'll remove that.

[Brent Fulgham]

Created attachment 295912 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 295912 [details]
Patch

Clearing flags on attachment: 295912

Committed r209226: <http://trac.webkit.org/changeset/209226>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.