[C++] g++.other/enum5.C

Tom Tromey tromey@redhat.com
Sun Jul 22 17:56:00 GMT 2001


>>>>> "Ben" == Ben Elliston <bje@redhat.com> writes:

>>>>> "Neil" == Neil Booth <neil@daikokuya.demon.co.uk> writes:
Neil> IMO we should xfail enum5.C.  After all, (almost) every xfail is a bug
Neil> we should be trying to fix eventually; I don't think anyone is in any
Neil> danger of forgetting that.

Ben> Except that this is not the intended purpose of XFAIL in DejaGnu.

In libgcj we decided to XFAIL anything we knew about.  Our reason was
that having some random number of FAILs made it hard for people to
know if libgcj was working.  "Is 72 the correct number of fails?  Or
is it 75?  I forget."  This happened all the time.

Now that we expect FAIL==0, it is easy to see if a port works or if a
change causes a regression.  On the minus side, we have to remember
that an XFAIL probably represents a real bug.  I think the next step
for us would be to make sure that each XFAIL has a corresponding PR.

In sum I'm glad we made this change.

Tom



More information about the Gcc-bugs mailing list