Re: [_Complex test 4/4 V3] _Complex tests in callfuncs.exp

[Yao Qi <yao at codesourcery dot com>]

On 05/24/2011 05:58 PM, Mark Kettenis wrote:
> Ah, I didn't realize that setup_kfail was part of dejagnu proper.
> Still I think that setup_kfail_for_target is what the behaviour of
> setup_kfail should be.  So dejagnu should be changed.
> 

IMO, it is a separate issue to change setup_kfail to what
setup_kfail_for_target like.

> I'd say you should just go with your previous version that uses
> setup_kfail, and submit the enhancement of setup_kfail to the dejagnu
> maintainers.  I believe you can link together bugs in bugzilla, so in
> the interim, you could use that feature to link together the various
> bug reports such that people can find the right one even if the bug ID
> that gets printed is the wrong one.
> 

There are various tricks in bugzilla to link different PRs together, but
it should not be recommended.  We don't know how long these bugs can be
fixed, so it is important to keep an *correct* PR number in KFAIL
message.  That is reason why KFAIL was invented, I think.

Despite of changes going to GDB or DejaGnu, the patch is almost the
same.  Only difference is setup_kfail vs. setup_kfail_for_target.

> Anyway, getting these tests in is more important than getting the
> linking to bug reports exactly right.  If other people think that

Now, we have the approach from which we can get tests running and get
the the correct bug report number.  As I said above, moving
setup_kfail_for_target to setup_kfail is a separate issue, and is a
little bit out of scope of GDB.

> setup_kfail_for_target is the way to go, feel free to ignore me.

-- 
Yao (éå)