Benoit Daloze

> what are the recommendations when to use such option? In short, for test suites it's typically worth trying. In general `--engine.Mode=latency` optimizes for latency at the expense of throughput....

* PR to use TRUFFLERUBYOPT=--engine.Mode=latency: https://github.com/rubocop/rubocop-ast/pull/241 * Issue about Parser::Lexer#advance: https://github.com/whitequark/parser/issues/871 With that let's close this issue.

That's odd, maybe a bug in the test harness? Is it using plain MSpec or something like RSpec + some extra files to define methods RSpec doesn't have?

> I'm not 100% sold on that JRuby and Truffleruby installation methods should do something "magic" and different from Ruby intallations. I would say it's also a problem for CRuby...

This is a bigger issue for `*-dev` than releases, so another option would be to only `rm -rf` or error out for dev builds. It could be an issue for...

TBH I'm really annoyed about this, I spent a large amount of time to try to fix this in a way that pleases @postmodern in chruby, spent time to discuss...

> > I am just curious, if this is such a critical issue, why can't truffleruby just workaround the problem? > > TruffleRuby already spent significant amount of effort to...

We saw that oj disables and re-enables the GC around parse_json, that's quite a hack 😅 From a quick look RB_GC_GUARD(str) should be used instead, and there is probably no...

Yes we were looking at `fast.c`

Regarding the string changing, how about `rb_str_dup()` (or just calling `dup`)? That uses copy-on-write and so should be faster as it does avoid the copy unless the string is actually...