memoizeN breaks on NET .10/11

[bartelink]

Description

dotnet/fsharp compiles this repo using the trunk compiler edition as a regression test
as of some point in the .NET 10 cycle this was determined to cause the memoization tests to fail
See #614

Its unclear what combo of the following represent the best resolution:

• change the compiler to make the test pass or reject the source
• a change to the source (did it rely on complier bug?)
• removing the SRTP-based overloads that failing tests use

Known Workarounds

only run tests on net8.0

Expected behavior

Tests pass on some commit in this repo

[bartelink]

Looks like @T-Gro's dotnet/fsharp#20247 resolves this

[gusty]

Thanks @bartelink
Labelling as an F# compiler bug as this is clearly a regression.
Usually a compiler issue should be opened and linked to this issue, which you did already with your last comment.
So, let's wait until dotnet/fsharp#20247 is merged and then we can close this.

[bartelink]

You're correct of course. However I didnt feel I had triaged to a significant enough degree, but wanted to have something to reference from #682 (comment) as I tried to unblock dotnet/fsharp#19785 (comment)
At this point I had not attempted to narrow down the root cause enough to be confident calling it 'clearly a regression' (unlike you I've not delved into SRTP to the point of considering that they can be recursive!).

[bartelink]

Marking resolved - compiler fix is in dotnet/fsharp main now

[wallymathieu]

Sounds really nice 😄