From @masak

<jnthn> m​: say "a\n" ~~ /^^\s*$$/
<camelia> rakudo-moar d69a87​: OUTPUT«Nil␤»
<jnthn> m​: say "a\r\n" ~~ /^^\s*$$/
<camelia> rakudo-moar d69a87​: OUTPUT«｢｣␤␤»
<jnthn> That's why Pod tables are bust on Windows.
<masak> jnthn​: that looks clearly wrong to me.
<masak> \r\n are whitespace.
<jnthn> masak​: That's not the problem
<jnthn> masak​: The problem is that it doesn't match, so then goes right one char
<jnthn> masak​: Then ^^ says "oh, I'm after a \r, that'll do", and $$
says "oh, I'm before a \n, that'll do"
<masak> oh!
<masak> ...wow.
<jnthn> Yeah... o.O
<jnthn> And given \r actually *is* a valid line sep on...older
mac...or something...
<jnthn> I don't immediately know what to do about it.
<jnthn> For now, it's easy to patch the Pod code, since it's already
dealing with full lines always
<lizmat> this is a similar problem with .lines
<jnthn> So its use of ^^ and $$ is kinda overkill; if you already know
you have a line, ^ and $ are good enough
<jnthn> But yeah, there is a deeper issue here
<masak> yes, \r was the newline marker on Mac OS 9.
<FROGGS[mobile]> I'd think that ^^ would "gobble" \r\n already
<jnthn> FROGGS[mobile]​: ^^ and $$ don't swallow
<jnthn> They're zero-width assertions
<masak> jnthn​: I think a reasonable expectation is that if you *have*
a single line, ^ and $ should act as ^^ and $$, respectively.
<jnthn> masak​: Yes, sounds sane. Question is how we get there ;)
<jnthn> Anyway, I'm going to do the wrokaround in Pod.pm so that we
don't have Pod-parsing bugs on Windows.
<jnthn> And the ^^$$ issue probably deserves an RT.
* masak submits one
<masak> does it make sense to ask "how does Perl 5 regexes handle this one?" ?
<jnthn> masak​: Maybe... :)

From @jnthn

On Fri Oct 03 03​:11​:45 2014, masak wrote​:

<jnthn> m​: say "a\n" ~~ /^^\s*$$/
<camelia> rakudo-moar d69a87​: OUTPUT«Nil␤»
<jnthn> m​: say "a\r\n" ~~ /^^\s*$$/
<camelia> rakudo-moar d69a87​: OUTPUT«｢｣␤␤»
<jnthn> That's why Pod tables are bust on Windows.
<masak> jnthn​: that looks clearly wrong to me.
<masak> \r\n are whitespace.
<jnthn> masak​: That's not the problem
<jnthn> masak​: The problem is that it doesn't match, so then goes
right one char
<jnthn> masak​: Then ^^ says "oh, I'm after a \r, that'll do", and $$
says "oh, I'm before a \n, that'll do"
<masak> oh!
<masak> ...wow.
<jnthn> Yeah... o.O

This is no longer a problem, thanks to \r\n being a single grapheme. Test coverage added in S05-metachars/line-anchors.t.

/jnthn

The RT System itself - Status changed from 'new' to 'open'

@jnthn - Status changed from 'open' to 'resolved'