Design Meeting Notes, 1/5/2024

[DanielRosenwasser]

Avoid Widening Returned/Yielded Types in Const Contexts

#56858

• Idea is to do a specific check when widening.
• Feels like an ad-hoc fix is not the right one?
• Why doesn't the const context give you the non-widening literal type in the first place?
• This is about const type paramerters - it's really more tied to ordering of when we decide to widen a function type.
  • Don't widen return types of function expressions #241
• Could take a few views:
  • This can be viewed as a suitable carve-out.
  • This can be duped to Don't widen return types of function expressions #241, in which case
    • search for a non-breaky more-ideal solution.
    • say "won't fix"
• We mostly feel like this is desirable - not like you need a ton of explanation of what's happening.
• Conclusion: we're mostly inclined to take the PR, barring any implementation problems.
  • Want tests for
    • Overloads
    • Async functions
    • Async yields

[DanielRosenwasser]

CC @Andarist

[christianalfoni]

As I understand from the conclusion you are inclined to fix it, but I have not seen updates in the referenced issue. Was just curious if you are still inclined to fix it? 😄

[Andarist]

This one has slipped my radar. I need to address the feedback for #56859 to get it landed. I don't think it fixes anyhow your issue though. This is solely about constness of those return statements/expressions when combined with const type parameters and not about "exactness"

[christianalfoni]

Oh, I missed the emphasis on const here. Sorry! But great I was able to put it back on the radar 😅 blip