Benoit Daloze

It feels like it's a bug in CRuby it's much faster at `gsub(/&/, MAP)` on BINARY than UTF-8 strings. In either case it's valid to just do a byte search...

One concern with it is it potentially discards the computed coderange, if it was known (unless it's CR_7BIT, that may be preserved across force_encoding's). I think it'd be good to...

It would be great indeed. In general we're trying to add truffleruby to all default & bundled gems CI (https://github.com/oracle/truffleruby/issues/2644) to avoid breakage and regressions like #452. Regarding the failures,...

It's not a correct fix, see https://bugs.ruby-lang.org/issues/21324#note-6

Thanks for the report. This needs to be addressed upstream, see https://github.com/oracle/truffleruby/issues/3118#issuecomment-1804194112 for the replacement.

It means the C extension is doing `RARRAY_AREF(ary, 0)` but ary is empty so that seems potentially a bug of the C extension, or somehow that array not having the...

> Will submit a PR to numo-narray once everything is passing. It would be helpful to already submit a PR with what you found. Notably that `RARRAY_LEN` could cause a...

https://github.com/rubygems/rubygems/commit/076aba8b1c25291a986f9c233cb80192adbe2fcf / https://github.com/rubygems/rubygems/pull/7719 seems to be what added this re-exec behavior for `require "bundler/setup"`, cc @deivid-rodriguez

I think it's probably fine to `exec` for binstubs generated by Bundler (because they cannot be passed ruby command line options AFAIK and they don't change CWD), but not for...

One more general thought on this is I expect (perhaps naively or wrongly) that `require 'bundler/setup'` basically just adds a bunch of entries to `$LOAD_PATH` and has no other visible...