Benoit Daloze

Thanks for the issue. Right, we should probably use the same logic as CRuby, even though that depends on the external encoding which feels brittle but that's the CRuby semantics...

Partial fix in https://github.com/truffleruby/truffleruby/pull/4001 merged, thanks, I suspect that fixes most cases in practice. Could you check if that makes it possible to remove the excludes in Prism?

Yes, thanks for adding that failing spec, it's useful to track that and also a good reason to keep this issue open.

> Recently I noticed logs for the CI run loading a 'close key' match cache, and it logged removing all the Rails gems. In theory that should be fine, because...

Copying it here for posterity. Looks fine, though the output is somewhat confusing with `Cache hit for` but actually using the cache from another key. Also `Print lockfile took 21.08...

Nice. Re `Remove struct RData deprecated by struct RTypedData` I know one usage in [concurrent-ruby](https://github.com/ruby-concurrency/concurrent-ruby/blob/master/ext/concurrent-ruby-ext/atomic_reference.c). I suppose the reason is TypedData was not available yet for the oldest supported version...

I see, maybe the `stty` backend could be enabled on MacOS/AArch64? That should work fine I'd think as it just shells out.

There is one: https://github.com/ruby/io-console/blob/master/lib/ffi/io/console/stty_console.rb https://github.com/ruby/io-console/blob/74c78afc243af6f3f0435afeb5ca3518287c7244/lib/ffi/io/console.rb#L57 is showing `ex` and not `$!` (the rescue above does not assign a variable), so not sure why the `stty` backend fails to be required,...

Ah right `stty` might not work in GitHub Actions, unsure. Ultimately this is an unexpected behavior difference between amd64 and arm64, so I guess the real solution is to support/fix...

I suspect it's the same issue as https://github.com/rubygems/rubygems/issues/8106