[Mlir-commits] [mlir] [mlir][emitc] Relax `hasSideEffect` rules for Var/Member (PR #145011)

[Mehdi Amini]

joker-eph wrote:

> , I'd say that we need a hasSideEffects which would be from the C point of view that could lead to UB/etc,, where reading variable from stack is generally assumed not do to so.

I don't quite get why UB matter for side-effects?
The whole discussion here is quite confusing to me, and the patch somehow suspicious (if you need an optimization with a special case, this does not seem like the right place to me for the fix, instead the transformation should become smarter).

https://github.com/llvm/llvm-project/pull/145011