[webkit-reviews] review requested: [Bug 81104] REGRESSION: getBBox returns incorrect results with empty containers : [Attachment 133663] Asking reviewers what they think about a different approach

[bugzilla-daemon at webkit.org]

Philip Rogers <pdr at google.com> has asked  for review:
Bug 81104: REGRESSION: getBBox returns incorrect results with empty containers
https://bugs.webkit.org/show_bug.cgi?id=81104

Attachment 133663: Asking reviewers what they think about a different approach
https://bugs.webkit.org/attachment.cgi?id=133663&action=review

------- Additional Comments from Philip Rogers <pdr at google.com>
I found that introducing an invalid objectBoundingBox is actually very noisy
and expensive: a lot of code, e.g., hit testing, ends up calling
objectBoundingBox() and needs to have guards added to check for the invalid
state. Furthermore, in most cases an empty bounding box (0,0,0,0) ends up being
the desired fallback anyway.

I've attached a speculative approach that goes a different direction: adding a
flag to just RenderSVGRoot and RenderSVGContainer for whether the contained
object bounding box is invalid. What do you think?