Use references in font loading code instead of pointers which can never be null

[Myles C. Maxfield]

Use references instead of pointers which can never be null

[Myles C. Maxfield]

Created attachment 430539 [details]
Patch

[Chris Dumez]

Comment on attachment 430539 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=430539&action=review

> Source/WebCore/css/CSSFontFaceSet.cpp:121
> +        Ref<CSSFontFace> face = CSSFontFace::create(*m_owningFontSelector.get(), nullptr, nullptr, true);

.get() should not be needed.

> Source/WebCore/css/FontFaceSet.cpp:169
> +    ASSERT(scriptExecutionContext);

I don't see how this is safe. FontFaceSet is an ActiveDOMObject and an ActiveDOMObject's scriptExecutionContext can definitely become null. I would imagine it is able to hit the assertion by:
1. Having the top frame get a FontFaceSet from a subframe
2. Remove that subframe from the document
3. Have the top frame call load() on that FontFaceSet

[Myles C. Maxfield]

(In reply to Chris Dumez from comment #2)
> Comment on attachment 430539 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=430539&action=review
> 
> > Source/WebCore/css/FontFaceSet.cpp:169
> > +    ASSERT(scriptExecutionContext);
> 
> I don't see how this is safe. FontFaceSet is an ActiveDOMObject and an
> ActiveDOMObject's scriptExecutionContext can definitely become null. I would
> imagine it is able to hit the assertion by:
> 1. Having the top frame get a FontFaceSet from a subframe
> 2. Remove that subframe from the document
> 3. Have the top frame call load() on that FontFaceSet

Aha! You're right. When I was reading ContextDestructionObserver I missed the implementation of the contextDestroyed() function.

[Myles C. Maxfield]

Created attachment 430547 [details]
Patch

[Myles C. Maxfield]

Created attachment 430548 [details]
Patch

[Chris Dumez]

Comment on attachment 430548 [details]
Patch

r=me

[EWS]

Committed r278466 (238486@main): <https://commits.webkit.org/238486@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 430548 [details].

[Radar WebKit Bug Importer]

<rdar://problem/78872787>