mruby-regexp: raise TypeError for a String argument to String#=~ - #6988

String#=~ defined by mruby-regexp delegates to re =~ self. When the argument is a
String, that dispatches back to String#=~, and the two recurse until the stack is
exhausted.

"abc" =~ "b"
# CRuby: TypeError (type mismatch: String given)
# mruby: SystemStackError

CRuby's rb_str_match() branches on the argument type: String raises TypeError,
Regexp matches, and anything else is delegated to that object's =~. Rejecting
everything that is not a Regexp would therefore be too strict -- it would break the
duck-typed argument that CRuby accepts. Only the String case needs an explicit guard;
delegation already produces CRuby-compatible behaviour for the rest.

Kernel#!~ is !(self =~ y), so "abc" !~ "b" now raises TypeError as well, matching
CRuby.

Tests for both the matching case and the TypeError are added to
mrbgems/mruby-regexp/test/regexp.rb.

One related divergence is left alone: "abc" =~ nil returns nil on CRuby (via
NilClass#=~) but raises NoMethodError on mruby. That is a NilClass question rather
than a String#=~ one, so it is out of scope here.

Summary by CodeRabbit

• Bug Fixes
  • Updated String#=~ to raise TypeError when passed a string, matching standard Ruby behavior.
  • Preserved pattern matching for supported inputs.
• Tests
  • Added coverage for matching, non-matching, and invalid string arguments to String#=~ and String#!~.