[Mlir-commits] [mlir] [mlir][Conversion] Generalize and fix crash in `reconcile-unrealized-casts` (PR #95700)

[Markus Böck]

https://github.com/zero9178 commented:

Looks like a nice simplification to me. Its a bit unfortunate the `populate` interface is being dropped as I don't think its uncommon for downstream users to include the pattern in their "mega dialect conversion pass" but what you suggested with just running the pass afterwards should be semantically equal.

I am a bit confused by the changes in some of the tests. How come these are caused by this change? My impression of the code was that this should be NFC except for the changed API basically

https://github.com/llvm/llvm-project/pull/95700