[LLVMdev] Jump Theading/GVN bug - moving discussion to llvm-dev

[Sanjoy Das]

> What would be implications of making dominates assert is given an
> unreachable block?  This seems like it would help isolate a lot of bugs.

I think that's a great idea.

Having said that, I think the major problem is with passes
*implicitly* assuming the acyclicity of the dominator relation (this
assumption is violated by graphs with unreachable blocks).  For
instance, such the assert you suggest would not have triggered in the
GVN case that started this thread.

-- Sanjoy