superreview requested: [Bug 423998] Fix repainting regression(s) and multiple invalidation bugs : [Attachment 313859] address review comments

Robert Longson <longsonr@gmail.com> has asked Robert O'Callahan (:roc) (reduced
activity March 18 to April 9) <roc@ocallahan.org> for superreview:
Bug 423998: Fix repainting regression(s) and multiple invalidation bugs
https://bugzilla.mozilla.org/show_bug.cgi?id=423998

Attachment 313859: address review comments
https://bugzilla.mozilla.org/attachment.cgi?id=313859&action=edit

------- Additional Comments from Robert Longson <longsonr@gmail.com>
(In reply to comment #17)

> nsSVGUtils.cpp: The DoUpdate changes are not necessary. Please revert them
for
> now. 

Done.

> In UpdateGraphic the NS_STATE_SVG_NONDISPLAY_CHILD check is not valid for
> all nsIFrame implementations. That bit is used for different things on
non-SVG
> frames. You should be able to just change the argument to be of type
> nsISVGChildFrame* and kill the QI.

Done, however GetStateBits is a method of nsIFrame so I still need a QI. Doing
it your way I need it from nsISVGChildFrame to nsIFrame rather than vice versa.


We could put guards into various functions using IsFrameOfType in some other
followup bug I guess.

> 
> My only other comment would be that the UpdateGlyphPositioning call in
> nsSVGTextFrame::NotifyGlyphMetricsChange seems quite undesirable. If that's
the
> way it was before then you're right, probably safest to just put it back for
> now.
> 

It's necessary to prevent some of the redraw regressions.