Tomas Grosup
Tomas Grosup
> Why does the trimmed assembly size change? Do we need to preserve nullness (and sig, opt) metadata? I would have thought no. It is new part of a signature...
> Why would Roslyn consume a trimmed F# assembly? You are right, will remove the IL attributes in case of aggressive trimming.
> Just a TLDR for the above linked issue (#16731) - I believe that `Option.ofObj` should only be valid for non-nullable ref types i.e `'T | null` - it is...
If this happens to be a really big pain point, I can have a look a #ifdef my way out of it - should be doable I think. But as...
Docs PR https://github.com/dotnet/docs/pull/40762/files#diff-5583236f005ad44753c6945d4312c09cbd439fb251f9a01db543048fe51d284bR1095 makes the suggestion an official and recommended default. A particular aspect I did not see in this thread is what this change causes to **diffs** : When...
Thanks @nojaf for visually demonstrating my concern, this is exactly it. The compiler codebase has many matches which go well above 20 cases, so this mimics a real world situation...
> It wasn't writing it to disk before either: > > https://github.com/dotnet/fsharp/blob/ddbaafd3533d7c2bc1fe8265938450fef4c6ad93/tests/service/Common.fs#L197 > > Which is probably the reason it didn't work for you during debug, because `Range.DebugCode` relies on...
In this branch, I locally added net8 to the list of target frameworks. Reason is to get a lot more nullness annotations, since BCL in its netstandard versions does not...
One new annoying thing noticed in recent addition I did (printfn %s supporting null). Imagine code like this: ```fsharp let process text = printfn "%s" text Call.Api_Which_Does_Not_Accept_Null(text) ``` It was...
## New(er) wave of pain points after some of the above have been resolved in the nullness impl ### Migration concerns - [ ] New need to explicitly annotate `when...