[llvm-commits] [llvm] r69946 - in /llvm/trunk: docs/LangRef.html lib/Transforms/Scalar/InstructionCombining.cpp lib/VMCore/Type.cpp

[Chris Lattner]

On Apr 26, 2009, at 12:41 AM, Sanjiv Gupta wrote:
> I think the assembler and dis-assembler never bothered to call  
> sequentialType::indexValid. This test case always used to work.
> Its only the clang and the opt that check the validity.
>
> But anyways there is a disconnect here between various tools and the  
> documentation. I think we should allow all types of integer indices  
> and change the documentation and code accordingly. I will do that.

Sounds great, thanks!

> Meanwhile, do you think we might have a similar case for  
> structType::indexValid. Adding a test case for that as well will be  
> a good idea.

I think it is reasonable to restrict struct indices to always be 32- 
bits.  They don't have anything to do with computable values on the  
target, they really are just indexes.

-Chris