This insane testcase triggers: ###!!! ASSERTION: Parent not consistent with exepectations: 'aOldParent == aFrame->GetParent()', file /Users/jruderman/central/layout/generic/nsBlockFrame.cpp, line 596 (The assertion message contains a typo, btw.)

This worksforme with an m-c build from rev 7c4cf0adb51f.

Still happens for me with m-c 02fe3defdfc5+.

Still worksforme with m-c b7d7aa6ac670, which is as recent as I can get right now. So what's different between our setups?

Still happens for me with m-c 064ff2a07928+. I'm using Mac OS X 10.5.x and Xcode 3.1. My mozconfig looks something like http://pastebin.mozilla.org/651718.

Jesse and I just debugged this on his machine. We're calling ReparentFrame from nsBlockFrame::ReflowDirtyLines (line 2179 in the latter), and we have aFrame->GetParent() == aNewParent already inside ReparentFrame. Since aNewParent->GetPrevInFlow() == aOldParent and aOldParent->GetNextInFlow() == aNewParent, the rest of the setup seems correct; it's just that the frame (a text frame in this case) already has the parent we want to reparent it to... roc, any idea what's up here?

It sounds like aFrame must have been moved to aOldParent's child list without its parent pointer being updated. Other than that, I have no idea.

WFM on trunk.

Crashtest: http://hg.mozilla.org/mozilla-central/rev/b64a7e4a35e2