[PATCH] Get all localization tests passing on linux!

Eric Fiselier eric at efcs.ca
Sun Aug 17 20:52:02 PDT 2014


Hi Jon,

Thanks for all the feedback.

> I'm with Dan on this... It seems like these 'fixes' are just lowering the
expectations of tests when testing against a GLIBC system. It's perfectly
appropriate to XFAIL them and let them fail if that is the case.

I've changed a fair amount of tests to XFAIL after Dan's advice. Are you
commenting on his comments or any tests in particular?

> If you're concerned about test coverage being lower because though there
are lots of assertions in a single test file, it only takes one failure to
effectively hide the results of the others, then I think it makes more
sense to find a way to split the test. That way the part that is XFAIL'd is
a bit more minimal.

I think that is the way I'll go but I'll get this patch in first. Hopefully
we can get a reasonable amount of linux coverage without lowering our
expectations of the tests.

> One more thing: You've marked a bunch of these tests as 'XFAIL: linux'.
Would 'XFAIL: glibc' be more appropriate?

Yes, it would. Currently we don't have a glibc available feature. Do you
know a nice way of detecting GLIBC during CMake configuration time or lit
configuration time?
For the sake of clarity it might be worth it to add a glibc available
feature whenever we are on linux but that is not any more correct.

Thanks for the feedback,
Eric




On Sun, Aug 17, 2014 at 9:45 PM, Jonathan Roelofs <jonathan at codesourcery.com
> wrote:

>
>
> On 8/17/14, 9:43 PM, Jonathan Roelofs wrote:
>
>>
>>
>> On 8/14/14, 9:43 PM, Dan Albert wrote:
>>
>>> I think I may have misled you when I said we should #ifdef the
>>> differences
>>> between glibc and Mac. If there are legitimate differences, we should
>>> #ifdef
>>> them. If glibc is wrong (it looks like it often is), we should just
>>> XFAIL the
>>> test and file a bug against glibc (or does that data come from an OS
>>> package?).
>>>
>> I'm with Dan on this... It seems like these 'fixes' are just lowering the
>> expectations of tests when testing against a GLIBC system. It's perfectly
>> appropriate to XFAIL them and let them fail if that is the case.
>>
>> If you're concerned about test coverage being lower because though there
>> are
>> lots of assertions in a single test file, it only takes one failure to
>> effectively hide the results of the others, then I think it makes more
>> sense to
>> find a way to split the test. That way the part that is XFAIL'd is a bit
>> more
>> minimal.
>>
>
> One more thing: You've marked a bunch of these tests as 'XFAIL: linux'.
> Would 'XFAIL: glibc' be more appropriate?
>
>
> Cheers,
>
> Jon
>
>
>>
>> Cheers,
>>
>> Jon
>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> cfe-commits mailing list
>>> cfe-commits at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>>
>>>
>>
> --
> Jon Roelofs
> jonathan at codesourcery.com
> CodeSourcery / Mentor Embedded
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140817/9bc73cce/attachment.html>


More information about the cfe-commits mailing list