[llvm-commits] Fwd: [llvm] r80947 - /llvm/trunk/test/Analysis/Profiling/profiling-tool-chain.ll

Andreas Neustifter astifter-llvm at gmx.at
Thu Dec 3 00:56:41 PST 2009


Sorry to ask, but would you mind trying this test again with the latest 
revisions? Unfortunatelly I can't reproduce this bug myself...

I haven't made any specific changes, but maybe it has resolved itself?

Thanks, Andi

On 09/11/2009 05:42 PM, Daniel Dunbar wrote:
> On Fri, Sep 11, 2009 at 1:55 AM, Andreas Neustifter
> <astifter-llvm at gmx.at>  wrote:
>> Hi Daniel,
>>
>> Andreas Neustifter wrote:
>>>
>>> Daniel Dunbar wrote:
>>>> I attached a tarball of the Output/ directory from the self-hosted
>>>> tests. If I compare to my local copy its clear what the problem is,
>>>> the OptEdgeProfCounts are wrong in the self hosted module. No clues as
>>>> to why yet. Do you have any idea what might be going on?
>>>
>>> No, not really. The initializer for OptEdgeProfCounters is wrong.
>>
>> Is there any progress on this?
>
> No, sorry.
>
>> What are the exact steps to produce this selfhosted build? I would like
>> to check on that this weekend...
>
> Well, if it reproduces easily then all you should need to do is:
>   1. build LLVM.
>   2. build llvm-gcc
>   3. build a new copy of LLVM configured to use your fresh llvm-gcc,
> and using that as the C/C++ compiler.
>   4. run make check (with those test portions enabled).
>
>   - Daniel
>



More information about the llvm-commits mailing list