-------- Forwarded Message -------- From: MenTaLguY <mental@...3...> To: Peter Moulder <Peter.Moulder@...38...> Subject: Re: r18847 introduces crash because of ... NULL Date: Wed, 11 Jun 2008 13:15:31 -0400

On Wed, 2008-06-11 at 23:43 +1000, Peter Moulder wrote:

(If you'll excuse me pointing out the obvious, it's probably not a good idea to deliberately introduce assertion failures into svn trunk.)

I did address the failures I knew of beforehand (modulo the duplicate(NULL) bits I had once FIXMEd but since forgotten about) and those which I had discovered through casual testing (I *did* try various permutations of saving, just not Save As Plain...).

However, there's ultimately no way around it -- if you add new assertions for things that can't be statically proven (either in absolute terms, or in practical terms due to e.g. the size of the codebase), there realistically *will* be some failures and you *have* to discover them empirically after introducing the assertions.  Absent a complete enough test suite, the only way to discover lingering failures in a reasonable amount of time is to release the change to a branch where there is a large pool of testers.

Incidentally, while the prohibition on NULL documents is new, the changes also caught at least one existing violation of an existing precondition:

https://bugs.launchpad.net/inkscape/+bug/239092

(I should have tested extensions though. That I've no excuse for.)

-mental