Benoit Daloze

> So **ruby-build** will only remove the prefix path if installing **Truffle Ruby**, and only if the directory already contains a previous installation, am I correct? Yes, or if the...

I don't want to take more time for this, obviously we have different opinions. IMHO if a developer installs anything into /usr/local, they are asking for trouble: * Not possible...

@radarek Sounds like a good way to reduce overhead. Of course, overhead for Fixnum#+ will always be whatever is around (or it should be billions i/s) :) `$MethodCall = MethodCall`...

> Every call of a form Foo.bar needs constant lookup. We can remove it using global variable, which is a little bit faster than constant lookup @radarek I missed that,...

@nateberkopec > I don't think so. Constants will always have to go through the lookup chain, because constants can be nested inside of each other, whereas globals will not. This...

@thedarkone For the memory model, speculating them as constants would still be fine as invalidation is thread-safe (it waits for every thread to throw the compiled code before continuing).

> This seems a limitation currently in MRI, but there is no reason why constants cannot have an inline cache too. Then the only difference is a cheap version check....

Agreed this seems a severe security issue, because it means any workflow using `actions/checkout` basically leaks the token to any process/action in that workflow which can just read it from...

In https://github.com/eregon/use-ruby-action/issues/7 we noticed an issue because using subtly different prebuilt Ruby binaries corrupts the cache. Being able to clear the cache when it's corrupted or contains errors would be...

One should include the OS and Ruby version in the cache key, as explained in https://github.com/ruby/setup-ruby#caching-bundle-install Interesting, maybe we shouldn't set `restore_keys:`, so the cache is invalidated if anything changes...