[perl #74488] [BUG] A closure retains the lexpad of the innermost call of the routine it's in in Rakudo

[Carl Mäsak]

# New Ticket Created by  "Carl Mäsak" 
# Please include the string:  [perl #74488]
# in the subject line of all future correspondence about this issue. 
# <URL: http://rt.perl.org/rt3/Ticket/Display.html?id=74488 >


<masak> I found the disastrous bug.
<masak> it took all of yesterday evening. and half of the night.
<jnthn> How disasterous is it?
<masak> jnthn: it's bad.
<jnthn> Damm.
<jnthn> Got a short test case?
<moritz_> do tell.
<jnthn> We've already been burnt by that.
<masak> rakudo: my $t = 0; sub foo { my @a; my &bar = { say +...@a };
unless $t++ { foo; push @a, "OH HAI"; say +...@a; bar; } }; foo
<p6eval> rakudo 842d2b: OUTPUT«1␤0␤»
<masak> alpha: my $t = 0; sub foo { my @a; my &bar = { say +...@a };
unless $t++ { foo; push @a, "OH HAI"; say +...@a; bar; } }; foo
<p6eval> alpha 30e0ed: OUTPUT«1␤1␤»
* masak submits rakudobug
<masak> I'll five you a minute or so :)
<moritz_> soo... &bar has some wrong lexical @a?
<masak> rakudo: my $t = 0; sub foo { my @a; my &bar = { say @a.WHICH
}; unless $t++ { foo; push @a, "OH HAI"; say @a.WHICH; bar; } }; foo
<p6eval> rakudo 842d2b: OUTPUT«87969168␤88746296␤»
<masak> yep.
<masak> alpha: my $t = 0; sub foo { my @a; my &bar = { say @a.WHICH };
unless $t++ { foo; push @a, "OH HAI"; say @a.WHICH; bar; } }; foo
<p6eval> alpha 30e0ed: OUTPUT«47405094140536␤47405094140536␤»
<masak> the @a in &bar gets bound to the @a in the foo call called by
foo, not the foo call called by the outermost scope.
<moritz_> maybe related to RT #73034, which also discusses variable
sharing in closures
* masak looks
<moritz_> that one is responsible for the man-or-boy test not working in master
<moritz_> Knuth++
<masak> I have no way to know whether mine is the same, so I'll go
ahead and submit it anyway.
<masak> they do look related.
<moritz_> maybe include a reference to that ticket
<masak> *nod*
<masak> until I can find a workaround, this bug blocks the conversion
of GGE to master.
<jnthn> pmichaud++ did at one point have a patch that got man or boy
passing again in master.
<moritz_> so why didn't he commit? did it break everything else?
<jnthn> Yes.
<masak> to heck with everything else!
<jnthn> rakudo: my $t = 0; sub foo { my @a; my &bar = pir::clone({ say
+...@a }); unless $t++ { foo; push @a, "OH HAI"; say +...@a; bar; } }; foo
<p6eval> rakudo 842d2b: OUTPUT«1␤1␤»
<masak> jnthn++
<masak> there's the workaround. (eww!) :)