I get this all the time while typing in the URL bar in Seamonkey: ###!!! ASSERTION: Never ran into the same child that we started from: 'sibling', file ../../../../mozilla/accessible/src/base/nsAccessible.cpp, line 825 Makes debugging suck. A lot. Please fix.

Needs s/DEBUG/DEBUG_A11Y

Uh... this doesn't fix the bug I filed. That fixes the e-mail I sent you, but that's not what this bug is about. Not that this is a bad change, mind you. ;)

Ginn or Evan, can you take a look at this?

like a dupe of bug 390414 when deleting the last character in text entry, because we fire delete event after the text node was deleted, we can't find the child in GetIndexInParent(), then get the assertion failed.

We can mark it as a dependency until we're sure.

Surkov, is this fixed by your patch in bug 393094?

(In reply to comment #7) > Surkov, is this fixed by your patch in bug 393094? > Really I can't reproduce the bug event without patch. I tried all seamonkeys I have.

I don't need Seamonkey. Just Firefox trunk, backspace in URL bar until last char is deleted.

Also, I was on Windows when I got the assertion.

I built trunk firefox, inabled ally by starting DOMi events viewer, removed all text from address bar by backspace and I see events from addressbar but I do not see any assertion.

I get an assertion, though it looks another one. Stack trace is: > accessibility.dll!nsHyperTextAccessible::DOMPointToHypertextOffset(nsIDOMNode * aNode=0x06505870, int aNodeOffset=0xffffffff, int * aHyperTextOffset=0x0012e724, nsIAccessible * * aFinalAccessible=0x0012e728, int aIsEndOffset=0x00000000) Line 659 + 0x12 C++ accessibility.dll!nsDocAccessible::CreateTextChangeEventForNode(nsIAccessible * aContainerAccessible=0x0650c3c4, nsIDOMNode * aChangeNode=0x06505870, nsIAccessible * aAccessibleForChangeNode=0x0650595c, int aIsInserting=0x00000000, int aIsAsynch=0x00000000) Line 1324 + 0x33 C++ accessibility.dll!nsDocAccessible::InvalidateCacheSubtree(nsIContent * aChild=0x06505850, unsigned int aChangeEventType=0x00000002) Line 1774 + 0x33 C++ accessibility.dll!nsAccessibilityService::InvalidateSubtreeFor(nsIPresShell * aShell=0x044f0cc8, nsIContent * aChangeContent=0x06505850, unsigned int aEvent=0x00000002) Line 1834 + 0x1f C++ gklayout.dll!nsGenericElement::doRemoveChildAt(unsigned int aIndex=0x00000000, int aNotify=0x00000001, nsIContent * aKid=0x06505850, nsIContent * aParent=0x053cd678, nsIDocument * aDocument=0x00dbd798, nsAttrAndChildArray & aChildArray={...}) Line 2721 C++ gklayout.dll!nsGenericElement::RemoveChildAt(unsigned int aIndex=0x00000000, int aNotify=0x00000001) Line 2692 + 0x2a C++ The problem is DOMPointToHypertextOffset() doesn't return an accessible when latest text node is removed (see http://lxr.mozilla.org/mozilla/source/accessible/src/html/nsHyperTextAccessible.cpp#661) and we get an assertion "Hypertext is reporting a different accessible for this node" (see http://lxr.mozilla.org/mozilla/source/accessible/src/base/nsDocAccessible.cpp#1354) Aaron, can you clarify this condition from DOMPointToHypertextOffset?

That's what is returned when the offset is at the very end of the text. Just change the assertion from NS_ASSERTION(changeAccessible == aAccessibleForChangeNode, to NS_ASSERTION(!changeAccessible || changeAccessible == aAccessibleForChangeNode,

Aaron's way

checked in