Benoit Daloze

1. No, it's not. all implementations with a JIT will likely be slower for tests/specs, but shouldn't be that much. In this case I guess it might be due to...

This is the check that fails: https://github.com/rubygems/rubygems/blob/18c579ea5c6d5f7325ede312bad0c2294b6fa4e1/bundler/lib/bundler/cli/exec.rb#L76 Compare to https://github.com/oracle/truffleruby/blob/b3faa0c24eb9a01922a0aa138d34b0ad137d7e41/bin/rake#L1 Maybe it's time to give up on our mixed Bash/Ruby `bin/*` executables.

You can use `--log.level=OFF`, that will effectively suppress the output. The warning is there for a reason, it means none of the standard library is available, no RubyGems, nothing. So...

I was thinking recently, maybe we can create the BooleanCastNode only if necessary? Like: ```java class SomeNode { BooleanCastNode booleanCastNode = null; boolean someMethodMustReturnBoolean() { Object result = ...; if...

Another idea is to have a general `executeBoolean(VirtualFrame) throws NotBooleanException` in addition to `execute(VirtualFrame)` and just try to call that optimistically, and switch to a BooleanCastNode when it fails. Downside...

Right, I guessed that as well. The issue is we capture the LexicalScope in the AST so: * it's hard to change it (e.g., it's stored as part of `ReadConstantWithLexicalScopeNode`)...

Interesting, RDoc uses yet another variant to check if it's valid Ruby syntax: https://github.com/ruby/rdoc/blob/2f7dfecdc3c90477e2f77e8cbadfba9cbe444814/lib/rdoc/markup/to_html.rb#L427-L434 This is what IRB uses now: https://github.com/ruby/irb/pull/134/files/5eb3ef3293a026e3f7de1711b8a399fe736bdb13#r702901657 So this issue is about supporting `return` in `BEGIN...

This also happens in `bundle exec rake --trace doc` of https://github.com/msgpack/msgpack-ruby/runs/6416790321?check_suite_focus=true: ``` $ bundle exec rake --trace doc ** Invoke doc (first_time) ** Execute doc [warn]: @param tag has unknown...

Some idea to fix this, we'd store the ReturnID in FrameDescriptor#getInfo(), and so when parsing the eval we could find the proper ReturnID to jump to. Semantics are like (i.e.,...

Thank you for the great report, it looks like pure-Ruby JSON parsing is slower with the String from `Zlib::GzipReader`. `Truffle::Ropes.flatten_rope()` seems a good hint that this is an issues with...