Grimy, has that changed? Just quickly trying to find info about that, and I see on regular-expressions.info/recursecapture.html it says "When Perl 5.10 [through to 5.18]'s regex engine enters recursion, all capturing groups appear as they have not participated in the match yet."

But that appears to have changed in 5.20

Ooh you’re right it appears to work on recent Perls

I had only tested on a 5.16

Yay

Also, I'm getting a very puzzling result from this perl bot on IRC (running v 5.24)

the result contains the value of \1 and \2 (a and b)

yeah

that’s what a regex with capturing group returns

and $& is set to the whole match, if you need that

The result implies "(?&b)" matches an empty string, so (?<=(?=$x|)[\s\S]) matches at the first opportunity, just after "b"

(?1) isn't matching empty there, at least

Found it :P I think "$x" isn't good enough of a non-empty contradiction for Perl

Wait, the opposite, Perl doesn't need those: (?=(?'a'[\s\S]*))(?'b'X(?=\k'a'\z)|(?<=(?=$x|(?&b)).)) works fine haha

er

I mean (?=(?'a'[\s\S]*))(?'b'X(?=\k'a'\z)|(?<=(?=(?&b)).))

Looks like Perl is doing something with "$x". Maybe some sort of optimization. Very puzzling. "x^" is fine though, so the expression can be touted for both Perl/PCRE now with that :)