[Mlir-commits] [mlir] [mlir][Analysis] Remove return value from `getBackwardSlice` (PR #163749)

[Han-Chung Wang]

hanhanW wrote:

I don't have context that when it'd fail. In the [previous revision](https://github.com/llvm/llvm-project/pull/140961#issue-3081414365), @wsmoses pointed that it could fail. Maybe @wsmoses can provide an example that helps the decision?

> This PR instead causes the analysis to return a failure, rather than crash in the cases it cannot compute the full slice

The PR was landed without tests. Therefore, I'm +1 on reverting the changes, although it's been like ~0.5 year. I think we all asked for tests in other PRs, but we did not get answers. See [question 0](https://github.com/llvm/llvm-project/pull/140961#pullrequestreview-2879240843) and [question 1](https://github.com/llvm/llvm-project/pull/142223#discussion_r2117533708).

The landed PR and a post fix are:
- https://github.com/llvm/llvm-project/pull/140961
- https://github.com/llvm/llvm-project/pull/142076

I'm more comfortable if we revert the changes and reland them with proper tests. It looks like the missing test is required in the discussion.

https://github.com/llvm/llvm-project/pull/163749