From @masak

<flussence> I broke nom last night!
<flussence> nom​: say ('aabaa' ~~ /\N+ b/).perl
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aabaa", from => 0, to
=> 3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<flussence> or I did, somehow
* moritz doesn't see the brokenness offhand
<flussence> nom​: say ('aaaaabaaaaa' ~~ /^(<[a..z]>*) b/).perl;
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aaaaabaaaaa", from =>
11, to => -3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<masak> huh!
<masak> that's... wrong.

(Note that the match fails, hence the 'to => -3'. There's clearly a
way for the regex to match, namely to map 'aaaaa' to the capturing
group and then match the 'b' literally.)

<flussence> it seems to not work with anything resembling a character class
<flussence> nom​: say ('aaaaabaaaaa' ~~ /^(\N*) b/).perl;
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aaaaabaaaaa", from =>
11, to => -3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<moritz> flussence​: it seems that the combincation of * and captures
somehow breaks
<flussence> nom​: say ('aaaaabaaaaa' ~~ /(\N*) b/).perl;
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aaaaabaaaaa", from =>
11, to => -3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<flussence> hm
<moritz> nom​: say so ('aaaaabaaaaa' ~~ /^(\N*) b/).perl;
<p6eval> nom 834d9d​: OUTPUT«Bool​::True␤»
<moritz> nom​: say so ('aaaaabaaaaa' ~~ /^(\N*) b/)
<p6eval> nom 834d9d​: OUTPUT«Bool​::False␤»
<masak> moritz​: is there an RT ticket for that yet?
<moritz> masak​: I'm not aware of any
* masak submits rakudobug

From @masak

On Tue Oct 04 02​:21​:54 2011, masak wrote​:

<flussence> I broke nom last night!
<flussence> nom​: say ('aabaa' ~~ /\N+ b/).perl
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aabaa", from => 0, to
=> 3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<flussence> or I did, somehow
* moritz doesn't see the brokenness offhand
<flussence> nom​: say ('aaaaabaaaaa' ~~ /^(<[a..z]>*) b/).perl;
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aaaaabaaaaa", from =>
11, to => -3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<masak> huh!
<masak> that's... wrong.

(Note that the match fails, hence the 'to => -3'. There's clearly a
way for the regex to match, namely to map 'aaaaa' to the capturing
group and then match the 'b' literally.)

<flussence> it seems to not work with anything resembling a character
class
<flussence> nom​: say ('aaaaabaaaaa' ~~ /^(\N*) b/).perl;
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aaaaabaaaaa", from =>
11, to => -3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<moritz> flussence​: it seems that the combincation of * and captures
somehow breaks
<flussence> nom​: say ('aaaaabaaaaa' ~~ /(\N*) b/).perl;
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aaaaabaaaaa", from =>
11, to => -3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<flussence> hm
<moritz> nom​: say so ('aaaaabaaaaa' ~~ /^(\N*) b/).perl;
<p6eval> nom 834d9d​: OUTPUT«Bool​::True␤»
<moritz> nom​: say so ('aaaaabaaaaa' ~~ /^(\N*) b/)
<p6eval> nom 834d9d​: OUTPUT«Bool​::False␤»
<masak> moritz​: is there an RT ticket for that yet?
<moritz> masak​: I'm not aware of any
* masak submits rakudobug

<flussence> nom​: say so ('aaaaabaaaaa' ~~ /^(a*) b/);
<p6eval> nom 7408d6​: OUTPUT«Bool​::True␤»
<flussence> it works fine for literals, not for char classes...
charset.t might fit
<moritz> nom​: say so ('aaaaabaaaaa' ~~ /^(.*) b/);
<p6eval> nom 7408d6​: OUTPUT«Bool​::False␤»
<moritz> nom​: say so ('aaaaabaaaaa' ~~ /^(<[a]>*) b/);
<p6eval> nom 7408d6​: OUTPUT«Bool​::True␤»
<moritz> nom​: say so ('aaaaabaaaaa' ~~ /^(<-[b]>*) b/);
<p6eval> nom 7408d6​: OUTPUT«Bool​::True␤»
<moritz> that kinda convinces me it's not char classes that are the
problem, but the fact that it needs to backtrack if that first thing
matches an 'a' too
<masak> aye.
<masak> I should've seen that too.
<masak> it's quantifiers and capturing groups that combine to create the
bug.
<moritz> nom​: say so 'aaaaaba' ~~ / ^ .* b/
<p6eval> nom 7408d6​: OUTPUT«Bool​::True␤»
<moritz> nom​: say so 'aaaaaba' ~~ / ^ (.*) b/
<p6eval> nom 7408d6​: OUTPUT«Bool​::False␤»
<moritz> nom​: say so 'aaaaaba' ~~ / ^ (.)* b/
<p6eval> nom 7408d6​: OUTPUT«Bool​::True␤»
* moritz agrees with masak's diagnosis
<flussence> nom​: say so ('ab' ~~ /^(.*) b/);
<p6eval> nom 7408d6​: OUTPUT«Bool​::False␤»
* masak adds this to the ticket

@masak - Status changed from 'new' to 'open'

From @jnthn

On Tue Oct 04 02​:21​:54 2011, masak wrote​:

<flussence> I broke nom last night!
<flussence> nom​: say ('aabaa' ~~ /\N+ b/).perl
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aabaa", from => 0, to
=> 3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<flussence> or I did, somehow
* moritz doesn't see the brokenness offhand
<flussence> nom​: say ('aaaaabaaaaa' ~~ /^(<[a..z]>*) b/).perl;
<p6eval> nom 834d9d​: OUTPUT«Match.perl(orig => "aaaaabaaaaa", from =>
11, to => -3, ast => Mu, list => ().list, hash => EnumMap.new())␤»
<masak> huh!
<masak> that's... wrong.

(Note that the match fails, hence the 'to => -3'. There's clearly a
way for the regex to match, namely to map 'aaaaa' to the capturing
group and then match the 'b' literally.)

Fixed this one a while back.

say ('aaaaabaaaaa' ~~ /^(<[a..z]>*) b/).perl;
Match.new(orig => "aaaaabaaaaa", from => 0, to => 6, ast => Any, list =>
(Match.new(orig => "aaaaabaaaaa", from => 0, to => 5, ast => Any, list
=> ().list, hash => EnumMap.new()),).list, hash => EnumMap.new())

And turned on various tests after doing so. So, resolving ticket.

/jnthn

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