Fix #5976: Don't assume T <:< (=>T).
#5981
Conversation
It's incorrect to assume `T <:< => T` since, taken as parameters, the erasure of the two types is different: `|T|` vs `Function1`. To compenate, we need to `widenExpr` in various places.
There was a problem hiding this comment.
Hello, and thank you for opening this PR! 🎉
All contributors have signed the CLA, thank you! ❤️
Commit Messages
We want to keep history, but for that to actually be useful we have
some rules on how to format our commit messages (relevant xkcd).
Please stick to these guidelines for commit messages:
- Separate subject from body with a blank line
- When fixing an issue, start your commit message with
Fix #<ISSUE-NBR>:- Limit the subject line to 72 characters
- Capitalize the subject line
- Do not end the subject line with a period
- Use the imperative mood in the subject line ("Add" instead of "Added")
- Wrap the body at 80 characters
- Use the body to explain what and why vs. how
adapted from https://chris.beams.io/posts/git-commit
Have an awesome day! ☀️
There was a problem hiding this comment.
Hello, and thank you for opening this PR! 🎉
All contributors have signed the CLA, thank you! ❤️
Commit Messages
We want to keep history, but for that to actually be useful we have
some rules on how to format our commit messages (relevant xkcd).
Please stick to these guidelines for commit messages:
- Separate subject from body with a blank line
- When fixing an issue, start your commit message with
Fix #<ISSUE-NBR>:- Limit the subject line to 72 characters
- Capitalize the subject line
- Do not end the subject line with a period
- Use the imperative mood in the subject line ("Add" instead of "Added")
- Wrap the body at 80 characters
- Use the body to explain what and why vs. how
adapted from https://chris.beams.io/posts/git-commit
Have an awesome day! ☀️
Align lub/glb with the changes to subtyping.
| @@ -603,7 +603,7 @@ trait Applications extends Compatibility { self: Typer with Dynamic => | |||
| * argument trees. | |||
| */ | |||
| class ApplicableToTreesDirectly(methRef: TermRef, targs: List[Type], args: List[Tree], resultType: Type)(implicit ctx: Context) extends ApplicableToTrees(methRef, targs, args, resultType)(ctx) { | |||
| override def argOK(arg: TypedArg, formal: Type): Boolean = argType(arg, formal) <:< formal | |||
| override def argOK(arg: TypedArg, formal: Type): Boolean = argType(arg, formal) <:< formal.widenExpr | |||
There was a problem hiding this comment.
Do we have some guarantee that the result of argType cannot be an ExprType here ? Even if we do, it seems more regular to widen both sides.
There was a problem hiding this comment.
The case we dropped in TypeComparer is where the RHS is an ExprType. => T was never a subtype of T.
There was a problem hiding this comment.
Right, but if both argType and formal are => T, then argOK used to return true but will return false now that we widen formal.
There was a problem hiding this comment.
I see. I believe argTypes are never ExprTypes. They might be TermRefs with an underlying ExprType. But we skip the ExprType when we do a widen on these. lub also does widen, not widenExpr. So I think we are good.
| @@ -98,7 +98,7 @@ object Implicits { | |||
| else if (mt.paramInfos.lengthCompare(1) == 0 && { | |||
| var formal = widenSingleton(mt.paramInfos.head) | |||
| if (approx) formal = wildApprox(formal) | |||
| ctx.test(implicit ctx => argType relaxed_<:< formal) | |||
| ctx.test(implicit ctx => argType relaxed_<:< formal.widenExpr) | |||
|
I'm wondering if we should introduce a |
I had a look at it. We have 8 occurrences of the pattern |
It's incorrect to assume
T <:< (=> T)since, taken as parameters,the erasure of the two types is different:
|T|vsFunction1.To compenate, we need to
widenExprin various places.