Adrián Montesinos González

@c-cube Not entirely, I pinned moonpool to simon/fix-35 and retried my tests, but the proposed fix kinda turns it into a heisenbug: While the above repro has looped for almost...

@polytypic I had not pinned picos. To check this isn't fixed already to some degree, I've just pinned all picos and moonpool packages to the main branch, reinstalled them on...

> Would it be possible to get access to that original test suite or something trimmed from it that triggers the issue? I'll try to share a more faithful repro...

@polytypic I've fixed the repro in https://github.com/amongonz/picos-issue to perform some "actual work" (copy 1 MiB from `/dev/zero` to `/dev/null` in chunks of up to 1 KiB) and added a script...

To compare with prior work, [F# added this feature not long ago][fsharp-syntax] using the `_.field` syntax. It's a bit more powerful there, as it allows chaining (e.g. `_.foo.bar`). I personally...

I found this suggestion when I was about to propose relaxing the unique-names rule to allow explicit sharing of fields, so ```F# [] type MyType = | Case1 of i1:int...

@Happypig375 I'm afraid your example wouldn't work, you can't overlap references and value-types in memory because the GC can't tell which type it's currently holding. Plus, you should respect alignment...

@Happypig375 You'd need cases to be of *unmanaged* types, which is even stricter. Otherwise, inner reference members could still overlap with non-references (or other misaligned references) and result in a...

@kerams I made a [comment suggesting explicit sharing of fields](https://github.com/fsharp/fslang-suggestions/issues/699#issuecomment-449115787) back in this thread which would address that. I also proposed it recently as a [separate suggestion](https://github.com/fsharp/fslang-suggestions/issues/1004) because I don't...

@T-Gro Fully agreed. I made the same proposal ages ago, twice, and I still think it's the only realistic path forward, but the allure of a more powerful solution lead...