[llvm-testresults] Broken buildbot: llvm-gcc-i386-linux-selfhost

[Duncan Sands]

Hi Daniel,

> Any chance you can run some quick checks on gcc11 to see if we can get
> some more information on these failures?

I will try to do that this weekend.  I was unable to reproduce on a local
machine.

> And/or hook someone else up with ssh access to investigate?

In general I am happy to give LLVM devs access to my accounts on these machines
(which are only used for running the buildbots) - send me your public ssh key if
you want access.

Ciao,

Duncan.

>
>   - Daniel
>
> On Fri, Sep 24, 2010 at 11:48 AM, Jakob Stoklund Olesen<stoklund at 2pi.dk>  wrote:
>>
>> On Sep 24, 2010, at 11:46 AM, Rafael Espindola wrote:
>>
>>>> I think I was wrong. I missed the stack traces.
>>>>
>>>> It looks like llvm-mc is crashing, producing a stack trace and truncated output.
>>>>
>>>> FileCheck is then complaining because the truncated output doesn't match. This caused weird error messages that I though looked like a miscompiled FileCheck.
>>>>
>>>> Note that many tests are passing -filetype=obj to llvm-mc, but none of them are failing in this bot.
>>>>
>>>> I suspect we are looking at a miscompilation of llvm-mc, or undefined behaviour in llvm-mc that LLVM exposes more than GCC did.
>>>
>>> Pain!
>>>
>>> I still think it is unlikely as tests for the parser are failing. I
>>> will try to revert 114678 and 114667 to see if that makes the bot
>>> happy.
>>
>> Thanks, Rafael.
>>
>> Just go ahead and put them back in if the bot stays red.
>>
>> /jakob
>>
>>
>> _______________________________________________
>> llvm-testresults mailing list
>> llvm-testresults at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-testresults
>>