This is the mail archive of the binutils@sourceware.org mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: New tests orphan2 and unknown2 fail in cris-elf


On Mon, Dec 19, 2005 at 04:32:51PM +0100, Hans-Peter Nilsson wrote:
> > Date: Mon, 19 Dec 2005 07:24:47 -0800
> > From: "H. J. Lu" <hjl@lucon.org>
> 
> > On Mon, Dec 19, 2005 at 02:48:39PM +0000, Paul Brook wrote:
> > > AFAIK that's always been the official policy.
> > > Obviously it's preferable to just fix the bug, but if that's not practical the 
> > > test should be XFAILed with a reference to the corresponding PR.
> > 
> > I don't see it. Pick one in
> > 
> > http://gcc.gnu.org/ml/gcc-testresults/2005-12/
> > 
> > and count how many FAILS in it.
> 
> Pointing out a flaw in following-through is not the same as
> saying it's not policy.  For each FAILs, IIUC they could be
> XFAILed as long as there's a PR for the issue, even if only seen
> as a testsuite regression.

It only shows that such a policy isn't enforceable. I don't think it
is a good idea to mark every fail in gcc testsuite as xfail nor I
believe it should be a gcc policy.

> 
> Besides, I bet none of those you might refer to are for *new*
> test-cases, known to fail when committed, which is the issue at
> hand.

Are you saying that if that testcase were checked in October, it would
be OK to fail?

I don't want to see any failures in binutils testsuite. But it doesn't
mean I'd like to mark them xfail unless noone is working on them or
they had never passed before. The only correct thing to do to a linker
regression is to fix it.


H.J.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]