From @masak

<cognominal> stuck anew with the same message; no return in the gather
though. https://gist.github.com/cognominal/5458754
<moritz> curious, I get "get_attr_str() not implemented in class 'Coroutine'"
<moritz> with cognominal's code
<cognominal> can we use gather in a recursive setting?
<cognominal> r​: sub a($n) { gather { take $n; take $n-1 if $n != 0
}}; say a(2);
<camelia> rakudo 37c995​: OUTPUT«2 1␤»
<cognominal> This simpler recursion case works
<masak> moritz​: I'm getting the same.
<masak> minimizing.
<moritz> cognominal​: iirc [Coke] has hit another limitation of
gather/take recentely; if you reuse the same gather/take block, the
first iterator has to be exhausted before the second one is started
<cognominal> arf
<masak> r​: sub p(*@​l) { gather { if @​l == 1 { take @​l } else {
@​l.shift; for p(@​l) -> @​p {} } } }; say p 1..4
<camelia> rakudo 37c995​: OUTPUT«get_attr_str() not implemented in
class 'Coroutine' [...]
<masak> there we go.
* masak submits rakudobug
<moritz> for p(@​l) -> @​p
<moritz> 'for' flattens, so if p(@​l) returns a list of arrays of
integers, that's what you get
<masak> moritz​: yes, I know.
<masak> moritz​: the internal error seems to happen before that, though.

From @usev6

I tried the evaluation from this bug report and didn't get that internal error anymore. There is still a type check error, though (which looks plausible to me)​:

$ perl6-m -e 'sub p(*@​l) { gather { if @​l == 1 { take @​l } else { @​l.shift; for p(@​l) -> @​p {} } } }; say p 1..4'
Type check failed in binding @​p; expected 'Positional' but got 'Int'
  in block at -e​:1

If I change @​p to a scalar container $p, the code does not die​:

$ perl6-m -e 'sub p(*@​l) { gather { if @​l == 1 { take @​l } else { @​l.shift; for p(@​l) -> $p {} } } }; say p 1..4; say "alive"'

alive

I'm not quite sure how a proper test for this ticket would look like. Just testing for X​::TypeCheck?

@usev6 - Status changed from 'new' to 'open'

From @coke

On Fri Jun 19 12​:57​:49 2015, bartolin@​gmx.de wrote​:

I tried the evaluation from this bug report and didn't get that
internal error anymore. There is still a type check error, though
(which looks plausible to me)​:

$ perl6-m -e 'sub p(*@​l) { gather { if @​l == 1 { take @​l } else {
@​l.shift; for p(@​l) -> @​p {} } } }; say p 1..4'
Type check failed in binding @​p; expected 'Positional' but got 'Int'
in block at -e​:1

If I change @​p to a scalar container $p, the code does not die​:

$ perl6-m -e 'sub p(*@​l) { gather { if @​l == 1 { take @​l } else {
@​l.shift; for p(@​l) -> $p {} } } }; say p 1..4; say "alive"'

alive

I'm not quite sure how a proper test for this ticket would look like.
Just testing for X​::TypeCheck?

Behavior changes in GLR​:
09​:05 < [Coke]> m​: sub p(*@​l) { gather { if @​l == 1 { take @​l } else { @​l.shift; for p(@​l) -> @​p {} } } }; say p 1..4
09​:05 < GLRelia> rakudo-moar 966096​: OUTPUT«␤»
09​:05 <+camelia> rakudo-moar fd1612​: OUTPUT«Type check failed in binding @​p; expected 'Positional' but got 'Int'␤ in block at
  /tmp/SqXtcgfOAz​:1␤ in block <unit> at /tmp/SqXtcgfOAz​:1␤␤»

--
Will "Coke" Coleda

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