Avoid infinite recursion when instantiating circular inline mapped generic tuple type #53522
Conversation
| // avoid infinite recursion, if the singleton is the type variable itself | ||
| // then we'd just get back here with the same arguments from within instantiateMappedType | ||
| if (singleton === typeVariable) { | ||
| return mappedType; | ||
| } |
There was a problem hiding this comment.
I don't really expect this to be the final solution to the problem. I think that it's likely to be possible to craft an extra test case with such an infinite recursion that would involve more than one layer between singleton and typeVariable. I wasn't yet able to manufacture such a test case though.
I'm not sure what's the best strategy to avoid such recursions in the compiler right now - there are different mechanisms like that in the codebase but I wasn't sure which one should be used here.
I imagine that we'd have to track what mapped types are being instantiated, and with what type arguments. This would allow us to bail out at some level. Unless perhaps - this is just a manifestation of the issue that lies elsewhere.
The other test case (with a type alias) works OK because it's able to remap that typeVariable to the alias' own type parameter~ since that alias has .mapper property set. So the mapper available here is the one that was combined here. That is not the case for an inline mapped type though since that has no .mapper
Andarist
deleted the
fix/infinite-recursion-when-instantiating-mapped-generic-tuple-type
branch
fixes #53458