Sema: Fix another SE-0110 issue [4.0] #9455
Conversation
|
@rudkx Can you review for 4.0? |
|
@swift-ci Please test |
|
LGTM. Thanks! |
|
Build failed |
|
Oops, looks like I should have run the tests locally first. Something that used to be ambiguous is no longer ambiguous -- might be an accidental bug fix but I need to investigate. I'll ping you when it's ready to review for real. |
The formal type of methods should be (Type) -> (Args...) -> (), not Type -> (Args...) -> (). This only matters in Swift 4 mode, where it was preventing the newly-added test case from type checking. Fixes <rdar://problem/31725325>.
slavapestov
force-pushed
the
se-0110-strikes-again-4.0
branch
from
a8bfa20 to
c49a06a
Compare
|
@rudkx The expected error that was no longer occurring in that test was something I added back when I implemented SE-0110, with a FIXME... and promptly forgot to ever investigate. It was the same bug. I've updated the test, this should be green now. There's no change to the logic, can I assume the LGTM still stands? |
|
@swift-ci Please test |
1 similar comment
|
@swift-ci Please test |
|
Build failed |
|
Yes, LGTM still stands - the change looks like the right thing to do. |
Description: Fixes a spot in the type checker where we were not wrapping a single-argument function input in a paren type, which caused type check failures in 4.0 mode.
Scope of the issue: Affects anyone migrating code from Swift 3 to Swift 4.
Risk: Low, this is a very narrow fix and the new type is more "correct".
Origination: Unmasked with the implementation of SE-0110 which is only enabled in Swift 4 mode, but the bug has been there all along.
Tested: New test case added.
Radar: rdar://problem/31725325.
Reviewed by: Mark Lacey