[Mlir-commits] [mlir] [mlir][vector] Fix round-trip parsing of operand-less vector.print (PR #231851)

[purnima shrivastava]

purnima-nlp wrote:

@matthias-springer @FedericoBruzzone 
Thanks a lot for the feedback

My understanding is that `vector.print` currently defaults to `PrintPunctuation::NewLine`.

As @matthias-springer suggested  wrapping operands in parentheses seems like a cleaner grammar-level solution, and I agree It would introduce a migration cost, as existing `.mlir` files, regression tests, and tools generating textual IR using `vector.print %x : type` would need to adopt the new syntax. I guess around 128 + tests mlir files depend on 'vector.print' .

@FedericoBruzzone, if the suggestion is to make bare `vector.print` a no-op or disallow it, wouldn't that also require updating existing IR? Making it a no-op would additionally change runtime behavior for programs relying on it to emit a newline, 
I'm not very sure if my correct . please guide me ...!
Is it better for this issue to be solved by changing semantics and migration?
@banach-space ?
Also I'm trying to deduce a solution that doesn't require migration and can fix it with minimal changes, I'm not sure if it exists. 

https://github.com/llvm/llvm-project/pull/231851