What I'm seeing is us failing to allocate objects, but for no good reason... :(

I actually hit one of these crashes on my Windows machine at home. Sadly I could not reproduce it.

Comment on attachment 9019529 [details] [diff] [review] Add instrumentation to figure out why Document.prototype creation is failing Review of attachment 9019529 [details] [diff] [review]: ----------------------------------------------------------------- ::: dom/bindings/Codegen.py @@ +2973,4 @@ > # if we don't need to create anything, why are we generating this? > assert needInterfaceObject or needInterfacePrototypeObject > > + def maybecrash(reason): Should this be "maybeCrash"? I don't know what the Python style is in this file. Also a name like "crashIfDocument" would be more descriptive. On the other hand, it would be less adaptable if you need to add yet more checks later. @@ +2974,5 @@ > assert needInterfaceObject or needInterfacePrototypeObject > > + def maybecrash(reason): > + if self.descriptor.name == "Document": > + return 'MOZ_CRASH("%s");\n' % reason Should the bug number go in here?

> Should this be "maybeCrash"? Then I couldn't use "maybeCrash" for the variable name. > Should the bug number go in here? Yeah, I guess that's a good idea. At this point I'm just querying for crash reports with ClearDocumentDepemdentSlots in the stack, but having the bug# in there is good.

Pushed by bzbarsky@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/f7910a01b533 Add instrumentation to figure out why Document.prototype creation is failing. r=mccr8