On Thu Oct 22 04:05:08 2009, masak wrote:
> In Rakudo 4c248e5c, constants can no longer be used inside of classes.
>
> $ perl6 -e'class A { constant B = 5; has $!x = B }; A.new'
> Null PMC access in isa()
> [...]
> $ perl6 -e'class A { constant B = 5; has $!x = A::B }; A.new'
> Null PMC access in isa()
> [...]
>
> I'm pretty sure this used to work before jnthn++'s merge. Which means
> we probably don't have a spectest for this.
>
> Sadly, it's not possible to refer to constants-defined-outside-a-class
> from within the class either.
>
> $ perl6 -e 'constant B = 5; class A { has $!x = B }; A.new'
> Lexical 'B' not found
>
> Here's a clean workaround, for those who want something like constants
> in a class:
>
> $ perl6 -e 'class A { sub B { 5 }; has $.x = B }; A.new.x.say'
> 5
<jnthn> Out of curiosity, does a lexical instead of a constant work?
<jnthn> How about an our-scoped variable?
* masak tries
<masak> jnthn: 'my' doesn't, 'our' works.
<masak> 'my' produces the same error as 'constant'. I'll add that to the ticket.
<jnthn> masak: OK. And since constants are package scoped, that probably gives
us a
pointer.
<masak> nod.
<jnthn> oh, my doesn't work?
<jnthn> Hmm
<jnthn> oh wait
<jnthn> I think constants became our scoped in the spec *after* I implemented
them. :-/
<jnthn> So they're probably still lexical.
<masak> aye.
<masak> that is consistent with the data.
<jnthn> *nod*
<jnthn> OK, so it looks like something to do with lexicals and attr
initializers.
<jnthn> Maybe we attach the initializer blocks to the wrong lexical scope.
