Karl Fogel wrote:
>
> julianfoad@btopenworld.com writes:
> > Perhaps I'm missing something, but I can't see when a crash would
> > ever be the expected behaviour. Well, only in an XFAIL test (which
> > is presumably a temporary measure), and in that case the worst I can
> > imagine happening is that people see XFAIL tests passing on Windows
> > and get confused about whether the corresponding bugs have actually
> > been fixed.
>
> Yes, you're missing something :-).
>
> No one on this thread is arguing that crashes should ever be expected
> behavior. The problem is that our test suite cannot reliably detect a
> crash -- hence, it has to look for other things that can only happen
> if there was *no* crash. If it fails to see those things, it can
> assume something went wrong.
>
> William/Philip posted a patch that would make it possible to detect
> crashes directly, but not for all platforms. My paragraph was about
> how it doesn't do us much good until it's portable, because we still
> have to write tests the same way we have been.
>
> So this isn't about XFAIL tests passing on Windows. It's about *any*
> claiming to pass on Windows when in fact the 'svn' client seg faulted
> under the hood. That's not good.
>
> Make more sense now?
Yes ...
I was thinking that it would surely be a good thing to add this _without removing_ the existing methods of detecting failure, because it would be an extra, more certain check. It is OK to have a bit more checking on certain platforms than on others. But I suppose the problem is that people would immediately start to rely on it _instead of_ the existing methods, which would result in worse checking (averaged across all platforms) than the status quo.
- Julian
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Apr 8 17:39:58 2003