[Mlir-commits] [mlir] [mlir][Transforms] Dialect conversion: Align handling of dropped values (PR #106760)

[Markus Böck]

https://github.com/zero9178 commented:

Is my intuition correct that this is rather difficult to test until the other PR relands but could then be tested by checking for a `() -> i32` `unrealized_conversion_cast` remaining in the IR (in the scenario showin in `@remove_all_ops`)?

Or could this maybe be tested right now be checking scenarios where previously two materializations would be created but now only one is?

https://github.com/llvm/llvm-project/pull/106760