Bug 204524 – [iOS] Copy assertions before iterating over them in _notifyAssertionsOfImminentSuspension

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit2

[Chris Dumez]

Copy assertions before iterating over them in _notifyAssertionsOfImminentSuspension.

[Chris Dumez]

<rdar://problem/57265830>

[Chris Dumez]

Created attachment 384166 [details]
Patch

[Alexey Proskuryakov]

Comment on attachment 384166 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=384166&action=review

> Source/WebKit/ChangeLog:13
> +        them. It is common for process assertions to get released when uiAssertionWillExpireImminently()
> +        gets called, which would remove them from the _assertionsNeedingBackgroundTask vector we were
> +        iterating on.

Is it possible to assertions to be added to the vector while iterating?

[Chris Dumez]

(In reply to Alexey Proskuryakov from comment #3)
> Comment on attachment 384166 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=384166&action=review
> 
> > Source/WebKit/ChangeLog:13
> > +        them. It is common for process assertions to get released when uiAssertionWillExpireImminently()
> > +        gets called, which would remove them from the _assertionsNeedingBackgroundTask vector we were
> > +        iterating on.
> 
> Is it possible to assertions to be added to the vector while iterating?

Not as far as I know, we're telling the clients that their assertion has expired and they're supposed to abort whatever they're doing (and release their assertion).
The "release their assertion" is logic I added recently, which is why this regressed.

[Alexey Proskuryakov]

Comment on attachment 384166 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=384166&action=review

> Source/WebKit/UIProcess/ios/ProcessAssertionIOS.mm:115
> +    Vector<WeakPtr<ProcessAndUIAssertion>> assertionsNeedingBackgroundTask = WTF::map(_assertionsNeedingBackgroundTask, [](auto* assertion) {

Seems worth explaining in a comment why we don't expect anything to be added to _assertionsNeedingBackgroundTask while iterating - or maybe even ASSERT at the end that there isn't anything new.

[Chris Dumez]

Committed r252811: <https://trac.webkit.org/changeset/252811>