[PATCH] D118297: [clang] add Diag -Wasm-volatile for implied volatile asm stmts

[James Y Knight via Phabricator via cfe-commits]

jyknight added a comment.

I agree that's the expected semantics. I think those semantics are unfortunate, but they're not gonna change. IMO it would've been better if you had to opt-in to "no side effects" via `__attribute__((const))` or so.

But I wonder why you think we should be discouraging people from writing "asm volatile" for things that are implicitly treated as having side-effects? It's quite non-obvious that adding an output argument to an existing asm statement changes whether it's treated as having side-effects, and I don't think it improves a codebase to make people lay out that trap more more frequently.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D118297/new/

https://reviews.llvm.org/D118297