From @masak

<masak> nom​: role R[​::T = role Q[​::S = role { method baz { say "OH
HAI" } }] { method bar { S.baz } }] { method foo { T.bar } };
R.new.foo
<p6eval> nom d59f09​: OUTPUT«===SORRY!===␤Can only use get_who on a
SixModelObject␤»
<masak> b​: role R[​::T = role Q[​::S = role { method baz { say "OH HAI"
} }] { method bar { S.baz } }] { method foo { T.bar } }; R.new.foo
<p6eval> b 1b7dd1​: OUTPUT«OH HAI␤»
<jnthn> oh wow
<jnthn> It explodes in the weirdest place
<jnthn> Well, more silliest than anything.
<jnthn> masak​: Guess you wanna know what's going on. Well, we
encounter the package. We then go and start parsing its signature. At
that point a $*PACKAGE is introduced, but nothing is bound to it yet.
When declaring Q, which should be installed in it, the package is thus
missing.
<jnthn> And it's nasty to fix because we tend to like to have seen the
traits before we go and create the type object (which the package
hangs off).
<jnthn> Because we want to know what the repr will be.
<jnthn> I guess really though Q should not be declared inside R
<jnthn> (e.g. it's not R​::Q)
<jnthn> As we're in the signature.
<jnthn> Not in the body.
<masak> right.
<jnthn> spectesting a fix
<masak> \o/
* masak submits rakudobug

From @jnthn

On Sat Oct 15 15​:36​:36 2011, masak wrote​:

<masak> nom​: role R[​::T = role Q[​::S = role { method baz { say "OH
HAI" } }] { method bar { S.baz } }] { method foo { T.bar } };
R.new.foo
<p6eval> nom d59f09​: OUTPUT«===SORRY!===␤Can only use get_who on a
SixModelObject␤»
<masak> b​: role R[​::T = role Q[​::S = role { method baz { say "OH HAI"
} }] { method bar { S.baz } }] { method foo { T.bar } }; R.new.foo
<p6eval> b 1b7dd1​: OUTPUT«OH HAI␤»
<jnthn> oh wow
<jnthn> It explodes in the weirdest place
<jnthn> Well, more silliest than anything.
<jnthn> masak​: Guess you wanna know what's going on. Well, we
encounter the package. We then go and start parsing its signature. At
that point a $*PACKAGE is introduced, but nothing is bound to it yet.
When declaring Q, which should be installed in it, the package is thus
missing.
<jnthn> And it's nasty to fix because we tend to like to have seen the
traits before we go and create the type object (which the package
hangs off).
<jnthn> Because we want to know what the repr will be.
<jnthn> I guess really though Q should not be declared inside R
<jnthn> (e.g. it's not R​::Q)
<jnthn> As we're in the signature.
<jnthn> Not in the body.
<masak> right.
<jnthn> spectesting a fix
<masak> \o/
* masak submits rakudobug

The fix went in; needs tests.

/jnthn

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

From @moritz

now tested in S14-roles/parameterized-type.t

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