PR17925, PATCH

[Stepan Dyatkovskiy]

ping
Stepan Dyatkovskiy wrote:
> I have updated fix:
> Introduced exceptions for
> * 'ret' instruction (we can ignore address space here)
> * 'load' instruction (we can't ignore address space in operands, but it we still can do it for instruction type itself)
>
> Now all tests are passed.
>
> -Stepan
>
> 14.11.2013, 16:38, "Stepan Dyatkovskiy" <stpworld at narod.ru>:
>> Please find bug description at:
>> http://llvm.org/bugs/show_bug.cgi?id=17925
>>
>> Sorry, I forget, that fixes should be published in llvm-commits, not in
>> bugzilla. So, below is fix description.
>>
>> Fix:
>> The are cases when we CAN treat pointers as integers, and places when we
>> CAN'T. Below is my proposal:
>> * We treat pointers as integers when we compare function formal arguments.
>> * Otherwise, we can't do that in general.
>> * We can add more exceptions, where address space affects nothing: for
>> example argument of "ret" instruction.
>> The patch is attached.
>>
>> Though, by now, it breaks 3 tests, but only since these tests checks
>> things that are wrong in general. We have to decide what to do with
>> these things:
>> 1. Can we always ignore address space of 'ret' type?
>> 2. Isn't it imprortant to tread 'load' instruction as different if they
>> accepts pointers with different address spaces?
>>
>> ,
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits