[Mlir-commits] [flang] [mlir] [mlir] Support DialectRegistry extension comparison (PR #101119)

[Nikhil Kalra]

nikalra wrote:

> Whether threading is enabling or not should not affect the invariants of the pass manager, the assert is very intentional in its current form.
> 
> Seems like an issue with detecting that some loading is a no-op?

It seems like this should be a better approach by allowing us to check if extensions have already been registered into the context (instead of conservatively assuming they aren't). But, this change touches 90+ extra files, so if there's a way to do this more cleanly I'm more than happy to investigate that further!

https://github.com/llvm/llvm-project/pull/101119