River Riddle

The LLVM conversion pass is currently a partial conversion, so no error is emitted if an operation is failed to convert.

The problem is that the conversion hasn't "failed" in the sense that you expect. A partial conversion allows for unconverted operations to exist the IR (by design). You are expecting...

Yes sorry, it is currently expected behavior, but not the desired behavior.

Thanks for reporting! Looks like an issue with mojo's handling of self-referential types. This is an issue with the mojo parser in general (the LSP is just a great fuzzer...

> Thanks for the answer! :) > > My confusion stemmed from the fact that I was expecting each instruction that is emitted to be emitted with a valid type....

This is from the type converter, which isn't provided with a source location.

A majority of the types come from block arguments, which don't have an exact location. Though I suppose we could use the location of the parent operation of the region...

@Mogball This doesn't work inside of a def either right now, it's possible that the implicitly defined variable isn't being added to the right scope.

Thanks a lot for the report! We discussed this a bit, but unfortunately decided to just remove tensorflow and related packages from the playground environment. Outside of configuration complications, there...

Thanks for filing this. This is essentially a duplicate of #15, in that right now top-level variables are treated a bit specially by the notebook environment. Ensuring that these variables...