https://github.com/erichkeane commented:
I'm a touch afraid this is going to be a wack-a-mole situation here with types, though if after 9 years, array is the only one we've found, maybe thats OK? An alternative approach might be to 'fixit' the type to `auto`/the deduced type and let it go from there. Presumably by the point we get to this point we've already diagnosed the mismatch? I actually wonder if the `replace with properly qualified auto` (ala our `replace bad type with int` everywhere else) might be a decent way forward/would allow the AST to progress better. I can see being OK with the 'wack a mole' approach with it/only doing it in cases we don't get to though (similar to what this is doing with invalid), but it would be nice if we coudl more gracefully recover from the two cases here. https://github.com/llvm/llvm-project/pull/210964 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
