[llvm-commits] [llvm-gcc-4.2] r45054 - /llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/930411-1.c

[Dale Johannesen]

On Dec 16, 2007, at 1:32 PM, Chris Lattner wrote:

>
> On Dec 16, 2007, at 1:29 PM, Dale Johannesen wrote:
>
>>>> URL: http://llvm.org/viewvc/llvm-project?rev=45054&view=rev
>>>> Log:
>>>> Suppress a valid warning for llvm.
>>>
>>> Having instcombine emit this warning seems like a really bad  
>>> thing to
>>> do.  Should we just stop the optimizer from emitting warnings like
>>> this?
>>>
>>> -Chris
>>
>> It's a useful thing to tell users about (users, unlike testsuites and
>> benchmarks,
>> are generally willing to change obviously wrong source).  Will it get
>> detected
>> anywhere else, particularly in the IPO case?
>
> I definitely agree, which is why I added it in the first place :).
> The problem is that it isn't hooked up with any of the other
> diagnostics machinery, so it isn't controllable and doesn't give
> decent location info (for example).
>
> I don't know that it really provides a good user experience.  In the
> LTO case with debug info, we could do better.

So what's your idea, detect it somewhere else?  Where were you thinking?