Replace http-parser with llhttp

[bryanp]

Builds on #639 to finish updating the response parser to use llhttp (using the ffi implementation for broader compatibility).

I updated one spec as skipped as llhttp does not return an error like http-parser does. Best I can tell llhttp exhibits the correct behavior, which was also @midnight-wonderer's take here: #639 (comment). Let me know if you disagree and I'll start a discussion with the authors of the llhttp c library.

Resolves #604, #622.

[bryanp]

@ixti or @tarcieri: how do you feel about dropping Ruby 2.4 support with this change? Ruby 2.4 is no longer maintained (2.5 will EOL soon as well). IMO this seems like a good opportunity to drop support for old rubies.

Right now https://github.com/metabahn/llhttp only supports Ruby 2.5 and up (explaining the failing checks above). I don't think there's any reason it couldn't support Ruby 2.4, but I'd prefer to only support maintained rubies. Let me know your thoughts.

[tarcieri]

I'm fine with bumping the minimum Ruby version to 2.5, or even 2.6 if 2.5 is about to be EOL.

[ixti]

Yeah with 5.x release I was hoping to bump the min ruby version.

[ixti]

I'm gonna merge #652 and then will rebase&merge this PR

[ixti]

Need to support Ruby 2.5 due to jRuby :((

[bryanp]

Speaking of jRuby, I'm trying to debug the failure above but can't reproduce locally. Have you seen this before?

exec: jrake: not found

Not sure why it's trying to use jrake—I don't see a reference to it in ffi-compiler.

[ixti]

It's not related to ffi-compiler. My branch with no other changes but ruby drop - fails on jruby as well o_O

[tarcieri]

@enebo @headius any idea what might be causing these jrake: not found errors which ostensibly appear to be coming out of ffi-compiler (even though that doesn't appear to directly reference jrake anywhere in its source code?)

[ixti]

@tarcieri I propose to temporarily exclude jRuby from the CI and deal with it later

[ixti]

@bryanp please rebase on master and I'll merge you PR. Will deal with jRuby later

[tarcieri]

lib/http/response/parser.rb:13:48: C: [Correctable] Style/HashSyntax: Use hash rockets syntax.
        @parser = LLHttp::Parser.new(@handler, type: :response)

@ixti it might be time to consider migrating to using the Ruby 1.9 syntax

[bryanp]

@ixti done

[headius]

what might be causing these jrake: not found errors

That is a new one to me... we have not shipped a "jrake" in years and years. There used to be a push (by others, not by us) to prefix every command installed in JRuby with a "j" like "jgem", "jrake", "jrails" and so on, so perhaps ffi-compiler got caught up in that?

[tarcieri]

@headius the confusing thing is I don't see jrake referenced anywhere in the ffi-compiler source code, nor do any of its dependencies (of which it only has a couple) stick out at me in particular as potential culprits

[ixti]

Thank you!

[bryanp]

@ixti Any idea when the next release might be pushed out?

[tarcieri]

See #659 and #660