[perl #82424] [BUG] Infinite recursion in parcel construction and assignment when the parcel contains the LHS in Rakudo

[Will Coleda via RT]

On Tue Jan 18 14:35:46 2011, masak wrote:
> <jnthn> rakudo: my $a = 1; my $b = 2; $a = ($a, $b); say $a.perl
> <p6eval> rakudo 388eed: OUTPUT«maximum recursion depth exceeded␤  in
> 'Block::count' at line 6077:CORE.setting␤  in 'Block::count' at line
> 6077:CORE.setting␤  in 'Any::join' at line 1␤  in 'List::perl' at line
> 2792:CORE.setting␤  in <anon> at line 2792:CORE.setting␤  in
> 'Any::join' at line 1␤  in 'List::perl' at …
> <jnthn> rakudo: my $a = 1; my $b = 2; $a ,= $b; say $a.perl
> <p6eval> rakudo 388eed: OUTPUT«maximum recursion depth exceeded [...]
> <jnthn> Bug.
> <masak> a known one, no?
> <jnthn> masak: I dunno, you're the human bug database. :P
> <masak> rakudo: my $a = 1; my $b = 2; $a ,= $b; say 'alive!'
> <p6eval> rakudo 388eed: OUTPUT«alive!␤»
> <masak> it's a bug in .perl -- how boring.
> * masak submits, Justin Case
> <jnthn> rakudo: my $a = 1; my $b = 2; $a ,= $b; $a.say
> <p6eval> rakudo 388eed:  ( no output )
> <jnthn> That looks...rather dubious too
> <masak> rakudo: my $a = 1; $a ,= 2; say $a.perl
> <p6eval> rakudo 388eed: OUTPUT«maximum recursion depth exceeded [...]
> <jnthn> rakudo: my $a = 1; my $b = 2; $a ,= $b; say ~$a
> <p6eval> rakudo 388eed: OUTPUT«maximum recursion depth exceeded [...]
> <jnthn> It's more than just .perl it seems
> <masak> aye.
> <masak> jnthn++
> * jnthn suspected so
> <masak> let's call it a bug in ,=
> <jnthn> :(
> <jnthn> wait
> * masak waits
> <jnthn> rakudo: my $a = 1; my $b = 2; $a = ($a, $b); say ~$a
> <p6eval> rakudo 388eed: OUTPUT«maximum recursion depth exceeded [...]
> <jnthn> Darn.
> <jnthn> It's a bug in parcel construction and assignment when the
> parcel contains the assignee it seems... :/
> <jnthn> Or something.
> <jnthn> Well, maybe.
> <jnthn> I note that stack trace has join in it, and the .perl one did
> too.
> <jnthn> rakudo: my $a = 1; my $b = 2; $a = ($a, $b); say $a.WHAT
> <p6eval> rakudo 388eed: OUTPUT«Seq()␤»
> <jnthn> rakudo: my $a = 1; my $b = 2; $a = ($a, $b); say $a[0]
> <p6eval> rakudo 388eed:  ( no output )
> <jnthn> rakudo: my $a = 1; my $b = 2; $a = ($a, $b); say $a[1]
> <p6eval> rakudo 388eed: OUTPUT«2␤»
> <jnthn> I wonder if something that should decontainerize is failing to
> do so.
> <jnthn> Though if so, it's hard to say what. Since l-value parcels
> surely shouldn't decontainerize on construction, or we break list
> assignment forms like ($a, $b) = 1,2
> * jnthn 's brain asplodes
> <jnthn> Well, file it. :)

Another data point - here's a working GLR version:

10:48 < TimToady> m: my $a = 1; my $b = 2; $a = (|$a, $b); say $a[0]
10:48 <+camelia> rakudo-moar 47ddca: OUTPUT«===SORRY!=== Error while compiling 
/tmp/gc6pUtxwbw␤Arg-flattening | is only valid in
                 an argument list␤at /tmp/gc6pUtxwbw:1␤------> my $a = 1; my $b 
= 2; $a = (|⏏$a, $b); say $a[0]␤»
10:48 < GLRelia> rakudo-moar fffbe7: OUTPUT«1␤»

-- 
Will "Coke" Coleda