Prism accepts double wildcard with no required fields in pattern match

[herwinw]

Slightly related to #2914

Prism parser the following code:

[1,2,3,4] => *a, *b

This results in a FindPatternNode with an empty list requireds.

MRI's parser rejects this syntax (I've tried this with Ruby 3.3.3 and a 3.4.0-dev of today):

ruby -e '[1,2,3,4] => *a, *b; p a; p b'
-e:1: syntax error, unexpected *
[1,2,3,4] => *a, *b; p a; p b
ruby: compile error (SyntaxError)

The behaviour is ambiguous: if we enable the prism parser in MRI, it turns out that the whole array is put into the right splat:

ruby --parser=prism -e '[1,2,3,4] => *a, *b; p a; p b'
ruby: warning: The compiler based on the Prism parser is currently experimental and compatibility with the compiler based on parse.y is not yet complete. Please report any issues you find on the `ruby/prism` issue tracker.
[]
[1, 2, 3, 4]

But the result a = [1, 2, 3, 4]; b = [] would have been just as valid. Other combinations like a = [1]; b = [2, 3, 4] would be valid too, but kind of random.

I would guess a FindPatternNode with no requireds should be a syntax error.

[herwinw]

The same thing can happen with an ArrayPatternNode:

[1,2,3,4] => *a, *b, 4

This too gets rejected by MRI

ruby -e '[1,2,3,4] => *a, *b, 4'
-e:1: syntax error, unexpected *
[1,2,3,4] => *a, *b, 4
ruby: compile error (SyntaxError)

Switching the parser to Prism (ruby 3.4-dev, the default included prism 0.30) results in a segfault.

In this case, MRI even rejects syntax that I would think is not ambiguous:

ruby -e '[1,2,3,4] => *a, 2, *b, 4'
-e:1: syntax error, unexpected ',', expecting end-of-input
[1,2,3,4] => *a, 2, *b, 4
ruby: compile error (SyntaxError)

This still results in a segfault when using --parser=prism

[kddnewton]

Thanks @herwinw! I'll get these fixed