[LLVMdev] ARMv5 Buildbot

Renato Golin renato.golin at linaro.org
Tue Feb 26 09:09:32 PST 2013


On 26 February 2013 16:14, David Blaikie <dblaikie at gmail.com> wrote:

> I'm not sure I understand your question. Fail tests continue to run so we
> find out if they start passing as seems to be the case here. If llvm is now
> free of whatever bugs caused these things to fail (ie: it's not just that
> these stopped failing on this hardware but in all environments the xfail is
> for) then they should just be un-xfailed.
>
Sorry, I got confused... starting over...

Some tests fail on ARMv7, others on ARMv5, and ExecutionEngine is old
enough (and never worked properly on ARM) for me to care much for it. My
question, and I think I have answered, is if it was possible to specify
sub-arch, and it is if it's in the triple (right?), so "XFAIL: armv5"
should do the trick if "armv5" is in the triple.

I confirm that the attached patch worked on my Panda and my laptop
(x86_64), but I think it won't work on the llvm-arm-linux bot, since it's
nor forcing "armv5" triple, but "arm-pc-linux" which won't match, though
the armv7 ones will not XFAIL on it, thus passing. I'll commit and see how
it behaves on the armv5 bot.

David / Xerxes / Galina,

Is it ok if we change the buildmaster to call that bot-group
"llvm-armv5-linux" or "llvm-arm-926"? So I can force the triple on the
buildmaster config?

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20130226/727833fe/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: armv5-tests.patch
Type: application/octet-stream
Size: 2674 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20130226/727833fe/attachment.obj>


More information about the llvm-dev mailing list