[webkit-reviews] review requested: [Bug 32839] REGRESSION (r52446) - Crash starting an OutlivePage load : [Attachment 46158] Patch minus the crashing

[bugzilla-daemon at webkit.org]

Nate Chapin <japhet at chromium.org> has asked  for review:
Bug 32839: REGRESSION (r52446) - Crash starting an OutlivePage load
https://bugs.webkit.org/show_bug.cgi?id=32839

Attachment 46158: Patch minus the crashing
https://bugs.webkit.org/attachment.cgi?id=46158&action=review

------- Additional Comments from Nate Chapin <japhet at chromium.org>
The additions to http://trac.webkit.org/changeset/52446 can be summarize as
such:
Request.h/cpp : In addition to the Frame, maintain RefPtrs to the Document and
DocumentLoader in the case of a request that can outlive the page.
DocumentLoader.cpp / ResourceLoader.cpp : Allow for the possibility that
Frame::page() and Frame::settings() may be null (as can now be the case if the
unload event is triggered by closing rather than navigating).
FrameLoader.h/cpp : Keep a count of the number of requests that may outlive the
page that refer to this FrameLoader, and don't allow m_documentLoader to be
nulled if that count is non-zero.

I'm not sure that this is a good solution, but it does appear to stop the
crashing.  It seems to me, though, that I've allowed the possibility that, when
an OutlivePage Request is executed, FrameLoader::m_documentLoader will not be
the DocumentLoader that originally associated with the Request.  Is that a Bad
Thing?