Nobuyoshi Nakada

Since fiddle is an external gem, so you have to send that part to https://github.com/ruby/fiddle. Also, it needs to be workable with previous ruby versions, and can't apply this blindly....

> The proposed solution is to make the call to `ruby_thread_has_gvl_p` unnecessary. Consider the case fiddle with this PR running with previous versions. The new fiddle may call `rb_thread_call_with_gvl` while...

> Yes, the code will need to continue using the syntax with the `ruby_thread_has_gvl_p()` check for a number of years. But in 2028, we will be able to remove the...

This seems to increase copies between `VALUE` and `rb_parser_string_t`. Right?

It is a common rule that a class/module that has no ractor-unshareable variables/constants is ractor-shareable, IIRC. `RbConfig::SIZEOF` and `RbConfig::LIMITS` are just the cases.

The [man page of macOS](https://developer.apple.com/library/archive/documentation/System/Conceptual/ManPages_iPhoneOS/man3/getpwent.3.html) says: > All of these routines are thread-safe. The `getpwent()`, `getpwnam()`, and `getpwuid()` routines return a pointer to a result managed by the system library...

Needs to update test/mkmf/test_pkg_config.rb. `-L` and `-l` in expectations would change.

At first, please propose a new feature at https://bugs.ruby-lang.org, not here. * It seems inconsistent with `Array#delete_at`. * It's questionable to return the default value for non-existent key. * The...

I think this kind of optimization is an area that compilers should take care of.

We need the discussion for new features at the BTS. Please file at https://bugs.ruby-lang.org/projects/ruby-master/issues/new.