pattern match bug with internal array

[HoneyryderChuck]

Given truffleruby's ruby 3 support, a new codepath was unlocked in httpx test suite which surfaced what it seems like a parser bug.

This should also be easily reproducible just by copying the line with the pattern match (I did it using irb and it crashed the console).

[eregon]

Pattern matching is not implemented yet, that's why it's a SyntaxError.
It's currently tracked as part of #2453.
The main issue is it's a huge effort to update the parser, with no easy solution.
We're trying to figure out a better long term solution to keep the parser up-to-date with reasonable efforts (instead of copying+adapting thousands of lines manually from JRuby or CRuby).

[eregon]

Does httpx support Ruby <= 2.6 which doesn't have pattern matching? It seems so from the gemspec & CI files.
What does it use to detect if pattern matching is supported?
The best way is probably to check one by trying to eval some pattern matching and if that's SyntaxError -> not supported.

[HoneyryderChuck]

Does httpx support Ruby <= 2.6

It does.

The detection is only relevant for tests, and is based on ruby version. For the library, the extensions are fairly safe even if pattern matching wouldn't be supported.

I guess I can use the eval as you suggested for now. It just feels a bit odd not going with the ruby version, as for other rubies it's a pretty reliable "feature discovery" tool (although I understand the issues related with the parser).

[eregon]

In general feature detection is much better than checking RUBY_VERSION (also applies to C APIs BTW), RUBY_VERSION should IMHO only be used when there is no other reasonable way (very rare).
Since TruffleRuby is released every 3 months, it's always possible some features of a CRuby version are not implemented, and so feature detection is important.

[HoneyryderChuck]

Sorry, I meant ruby version is pretty reliable if I can't do feature detection (i.e. defined?(NewClass) or .respond_to?(:fancy_new_method)) :-) . I guess there isn't a great way to feature-detect new syntax, besides your suggestion, so I'll take it.

[HoneyryderChuck]

Applied the patch, should this be kept open?

[eregon]

Let's close this, we have other issues to track pattern matching support, and until then the SyntaxError is expected.