@rhenium done

Could you add a test case for this class, too?

@rhenium added test.

FWIW I couldn't make the config frozen when initialized, as that interferes with .dup, which should return an unfrozen copy. Not sure whether it's worth it to force this though, as it seems to be a breaking change, and I have no particular use-case for it either.

FWIW I couldn't make the config frozen when initialized, as that interferes with .dup, which should return an unfrozen copy. Not sure whether it's worth it to force this though, as it seems to be a breaking change, and I have no particular use-case for it either.

I think it's acceptable that a #dup returns a frozen object since it is what some core classes like Interger do in Ruby >= 2.4.

I think making OpenSSL::Config frozen by default makes sense. At the same time, I think this class is mostly only useful when working with OpenSSL::X509::ExtensionFactory, and it wouldn't be an interesting object to share between ractors. Either way, such a change should be out of scope of this PR.

managed to make it frozen when initialized. turns out that .parse wasn't calling config_initialize as I thought, and that got me blindsided.

It seems strange to me that Config.new returns a frozen object and #dup returns a non-frozen object. Shouldn't the latter return a frozen object, too?

I think making it shareable between ractors and making it frozen by default are two separate changes even if both make sense. Can you split them into two commits?

Indeed, .dup should return a frozen config, fixed.

Can you split them into two commits?

done.