[parser] Fix bug when printing type builder errors #6817
Conversation
The type index from the TypeBuilder error was mapped to a file location incorrectly, resulting in an assertion failure. Fixes #6816.
| @@ -34,7 +34,7 @@ Result<> parseTypeDefs( | |||
| if (auto* err = built.getError()) { | |||
| std::stringstream msg; | |||
| msg << "invalid type: " << err->reason; | |||
| return ctx.in.err(decls.typeDefs[err->index].pos, msg.str()); | |||
| return ctx.in.err(decls.subtypeDefs[err->index].pos, msg.str()); | |||
There was a problem hiding this comment.
Why is this correct? I'm not sure how I would have known to think this was wrong. We are in parseTypeDefs, so shouldn't the typeDefs be relevant..? We even loop over typeDefs on line 29.
There was a problem hiding this comment.
Each entry in typeDefs corresponds to a single recursion group (which may consist of a single top-level type definition without (rec...), whereas each entry in subtypeDefs corresponds to a single type definition, which may or may not actually contain (sub ...).
Better names would perhaps be recTypeDefs and typeDefs rather than typeDefs and subtypeDefs. Do you think it's worth updating those names here or in a separate PR?
There was a problem hiding this comment.
I see, thanks. Either way sgtm. lgtm as is if you prefer separately.
The type index from the TypeBuilder error was mapped to a file location
incorrectly, resulting in an assertion failure.
Fixes #6816.