Add tests to regexp_encoding_option_mismatch

[thdaraujo]

(related to #2667)

The error can be seen in CRuby with parse.y enabled:

ruby --parser=parse.y -e "p /Ȃ/e.encoding"

#=> -e:1: regexp encoding option 'e' differs from source encoding 'UTF-8'
ruby: compile error (SyntaxError)

It could also be seen on Ruby 3.4.7 with Prism enabled:

ruby -e "p /Ȃ/e.encoding"
-e: -e:1: syntax error found (SyntaxError)
#=> > 1 | p /Ȃ/e.encoding
        |       ^ regexp encoding option 'e' differs from source encoding 'UTF-8'

And by parsing directly, you get the error too:

bin/parse -e "p /Ȃ/e.encoding"

Errors:
[#<Prism::ParseError @type=:regexp_encoding_option_mismatch @message="regexp encoding option 'e' differs from source encoding 'UTF-8'" @location=#<Prism::Location @start_offset=7 @length=1 start_line=1> @level=:syntax>]

Looking at the CRuby and Prism implementations, both are doing the same checks.

But the CRuby code is only enabled for Ripper (with the #ifndef RIPPER).

So maybe the problem is that Ruby + Prism is raising the Syntax error, but it should only be raised when using Ripper, and not during normal execution? But that would contradict the original issue description, which expects the syntax error to be raised.

I might be missing some context here about what the right behaviour should be, but I went ahead and added some tests to make sure the current behavior is the expected one. Hope this makes sense!

Thanks!

[thdaraujo]

should this test be added to prism/encoding/regular_expression_encoding_test.rb instead?

[kddnewton]

It's fine here.

[Earlopain]

Looks like this may have already been implemented in #2486. It's a bit weird since the issue was openend a month after that PR got merged.

Ripper shouldn't matter here, what's important is that ruby -c test.rb and ruby --parser=parse.y -c test.rb report the same thing (I know very little about how the original parser is actually implemented). prism and parse.y seem to agree, so all seems good?

[thdaraujo]

@Earlopain yeah that's what I was thinking too. Maybe we can close the issue.

Do we care about adding these tests, or should I just close this PR?

Thanks!

[kddnewton]

Hmm... it would appear it's as you say, and the original can be closed