Re: svn rev 4138: FAIL (i686-pc-linux-gnu shared)

"B. W. Fitzpatrick" <fitz@red-bean.com> writes:
> > OK, but shouldn't a test that supposed to fail be considered a test that
> > passes :)
>
> Oh *gawd* please don't start this discussion again. This dead horse
> was thoroughly beaten a while back.

:-) Well, let's explain it:

"XFAIL" is traditional for a test that reproduces an undesirable
behavior. We didn't invent the term; we chose it because a lot of
people (though not everyone) already know what it means.

It allows you to have a regression test running before the bug is
fixed. After the bug is fixed, the test fails to fail, so you change
it to a PASS. Now if the bug ever reappears, you'll get a real
failure (FAIL, not XFAIL).

The reason you don't want to just call it a PASS in the first place is
that if you accidentally fix the bug, you don't want to see a test
"failing", rather, you want to see that a test is unexpectedly
passing. XFAIL allows this distinction.

-Karl

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Dec 17 17:42:37 2002