=?UTF-8?B?QnJhbmtvIMSMaWJlag==?= <brane@xbc.nu> writes:
> Please, before anyone starts fixing this, write some tests that tickle
> the problem. In fact, this should be SOP for bug fixes:
>
> * write a test that exposes the defect
> * fix the bug
> * check that the test now passes
>
> We'd need some sort of XFAIL mechanism, though, so that we can mark
> tests that are expected to fail. The rule for commits would then be that
> XFAILs are allowed, but FAILs are not. I'll look into this.
Yeah, I was kinda against the whole XFAIL notion at first, but lately
there have been several tests that I've wanted to commit up as
expected failures that I had planned to fix immediately thereafter,
but had to instead #if-0 out.
It occurs to me that if for every bug we encountered, we first wrote
an XFAIL test to expose it, that might be a way to encourage
volunteers to work on fixing the bugs themselves. I mean, if we've
already provided a) the recipe for exposing the bug, and b) the
testing framework for showing when a bug gets fixed (goes from XFAIL
to SUCCESS), that's two fewer things a volunteer has to screw with.
For what it's worth, I'm +1 on XFAIL.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Aug 21 21:52:57 2002