Fix: detect error when trying to use unbound generic type in instance var #6128
Conversation
|
I think this fix is fine, but I think the error message is weird to me. Shouldn't we print the long |
|
I won't be coding anymore for a while, so if somebody wants to improve this, either modify it after this PR, in this PR or send a separate PR. |
|
Come on, just please merge this for 0.25.0. A guy just stumbled upon this error and nobody seems to notice it's a duplicate (for like a thousand times). A wise guy once said to me that it's better to make small progress instead of wanting to attain perfection. I believe this error message is better than an internal compiler error. Take it or leave it. |
|
It's too late for 0.25.0, it's already been tagged and built, just waiting to be signed and released. I'm sure we can get this in for 0.25.1. |
Fixes #5383
Sorry, this one is a bit hard to explain, why the fix works. But basically, an uninstantiated generic type should only be instantiated with available free vars, not all the time (in the new test case,
Tshouldn't be found because we are trying to find instance vars ofFoo, and soGen(T)shouldn't be a valid type for an instance var ofFoo(it could be valid for an instance var ofGen(T), but that's not the case).