Re: PATCH: error reading variable: value has been optimized out

[Jan Kratochvil <jan.kratochvil@redhat.com>]

On Fri, 14 Sep 2012 23:39:37 +0200, Andrew Burgess wrote:
> On 14/09/2012 8:02 PM, Tom Tromey wrote:
> >Andrew> +	# If we ever fix gdb so this passes we should delete the
> >Andrew> +	# unsupported case below.
> >Andrew> +	xpass $test
> >
> >Andrew> +	unsupported $test
> >
> >I think these tests should 'pass' if they generate the correct output,
> >and 'kfail' otherwise.

I agree with Tom here.

There are ongoing disputes how to mark testsuite results as it is more
a matter of opinion, so I do not want much to argue about anything.
Also I am not sure if I wrote it below so that it can be well understood.


> But if they kfail then they need a defect ID, it seems rather
> pointless to raise a defect for unsupported behaviour we have no
> intention to fix.

Not everything may be perfect but it is the current practice to do it that
way.


> Surely xpass is the right choice.

'xpass' is OK.  But that 'pass' is not right, we know GDB could behave better
(*). 'kfail' is for the 'pass' state more appropriate as we know GDB has
a (theoretical) problem.

(*) It is not important here whether it is worth the effort which it is not.

XPASS
     A pleasant kind of failure: a test was expected to fail, but
     succeeded. This may indicate progress; inspect the test case to
     determine whether you should amend it to stop expecting failure.


> If the test ever does pass I hope
> the unexpected pass result will draw attention to the test so the
> "unsupported" (or kfail) can be removed.  If it was just a pass then
> there's a chance the test could be fixed then regress and we'd never
> notice as the unsupported (or kfail) would never have been removed.

'unsupported' should not apply in any of these cases as it is about the testing
environment.  But all these testcases do not depend on anything in system.

UNSUPPORTED
     A test depends on a conditionally available feature that does not
     exist (in the configured testing environment).  For example, you
     can use this outcome to report on a test case that does not work
     on a particular target because its operating system support does
     not include a required subroutine.


Thanks,
Jan