From @masak

<masak> r​: role R[​::T] {}; class B {}; class C does R[B] {}; say "alive"
<p6eval> rakudo 35b2a9​: OUTPUT«alive␤»
<masak> r​: role R[​::T] {}; class B {}; class C is R[B] {}
<p6eval> rakudo 35b2a9​: OUTPUT«===SORRY!===␤Cannot call
'trait_mod​:<is>'; none of these signatures match​: [...]
<masak> std​: role R[​::T] {}; class B {}; class C is R[B] {}
<p6eval> std 7deb9d7​: OUTPUT«ok 00​:00 45m␤»
<masak> I realize that Rakudo doesn't have much to work with here...
<masak> (it's essentially the same situation as earlier, with
trait_mod​:<is> doing all the heavy lifting, and we shouldn't muck too
much with that)
<masak> ...but I had to realize all on my own that I had used 'is'
where I shoulda used 'does'.
<masak> jnthn​: is that worth a rakudobug, or should I just write code
correctly from the beginning? :P
<jnthn> masak​: um, I thought we should pun the role into a class and
inherit from that.
<jnthn> masak​: So file a "prolly shunt fail at all" bug :)
* masak files a "prolly shunt fail at all" rakudobug

From @Mouq

On Sun Jan 20 14​:06​:41 2013, masak wrote​:

<masak> r​: role R[​::T] {}; class B {}; class C does R[B] {}; say "alive"
<p6eval> rakudo 35b2a9​: OUTPUT«alive␤»
<masak> r​: role R[​::T] {}; class B {}; class C is R[B] {}
<p6eval> rakudo 35b2a9​: OUTPUT«===SORRY!===␤Cannot call
'trait_mod​:<is>'; none of these signatures match​: [...]

The problem is that Rakudo parses "C is R[B]" as trait_mod​:<is>(C, R, [B]), rather than trait_mod​:<is>(C, R[B]).

<masak> std​: role R[​::T] {}; class B {}; class C is R[B] {}
<p6eval> std 7deb9d7​: OUTPUT«ok 00​:00 45m␤»

STD parses that the same way as Rakudo, but I don't think we should clone the action behavior of <typename> (which handles "does R[B]") inside the action method of trait_mod​:is.

I'm not sure what a solution is at this time.

~Mouq

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