From @LLFourn

m​: say my Any​:D $a = Nil # (Any​:D)

that should be an error.

http://irclog.perlgeek.de/perl6/2016-04-22#i_12377287

From @zoffixznet

Still present in 405519

<Zoffix> m​: say my Any​:D $a = Any
<camelia> rakudo-moar 405519​: OUTPUT«Type check failed in assignment to $a; expected Any​:D but got Any (Any)␤ in block <unit> at <tmp> line 1␤␤»
<Zoffix> m​: say my Any​:D $a = Nil
<camelia> rakudo-moar 405519​: OUTPUT«(Any​:D)␤»

From @skids

rakudo PR #​829 should fix this on moarvm. JVM and JS fixes may also be needed.

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

From @gfldex

16​:02 < TimToady> I think my Int​:D $a = 42 should initialize as soon as possible, and never allow assignment of anything that ~~ Nil
16​:02 < TimToady> bare "my Int​:d $a;" should be illegal at compile time
16​:02 < TimToady> the invariant is what's wanted, and that's what we should maintain
16​:04 < TimToady> this is one of those spots where we should torment the implementor on behalf of the user, I suspect

16​:10 < TimToady> so when the symbol is introduced, we can temporarily poke an Int​:D into it, but as soon as the "my" executes and sets the value the first time, it should never be reset to Int​:D except by re-entry
16​:11 < TimToady> if we know the value at compile time, we can even just initialize it to 42 when the symbol is created, even before the 'my' executes

From @skids

On Fri Jul 22 15​:34​:49 2016, bri@​abrij.org wrote​:

rakudo PR #​829 should fix this on moarvm. JVM and JS fixes may also
be needed.

I added some error awesomeization to that PR so that now​:

$ perl6 -e 'my Int​:D @​a; @​a[0] = Nil; @​a.say;'
Type check failed in assignment to @​a; expected type Int​:D cannot be itself (perhaps Nil was assigned to a :D which had no default?)
  in block <unit> at -e line 1

$ perl6 -e 'my Int​:D @​a; @​a[0] = Int​:D; @​a.say;'
Type check failed in assignment to @​a; expected type Int​:D cannot be itself (perhaps Nil was assigned to a :D which had no default?)
  in block <unit> at -e line 1

$ perl6 -e 'my Int​:D $a = 42; $a := Nil; $a.say;'
Type check failed in binding; expected Int​:D but got Nil (Nil)
  in block <unit> at -e line 1

$ perl6 -e 'my Int​:D $a = 42; $a := Int​:D; $a.say;'
Type check failed in binding; expected type Int​:D cannot be itself
  in block <unit> at -e line 1

And though there is an awesomeization for sub return types it does not work,
because the X is being fed the raw type not the smilied typed.

$ perl6 -e '{ sub f() of Str​:D { Str​:D }; f(); CATCH { $_.expected.WHAT.say }}'
(Str)
Type check failed for return value; expected Str but got Str​:D (Str​:D)
  in sub f at -e line 1
  in block <unit> at -e line 1

...that is a separate issue so, not touching it.

Another thing I'm not touching is this​:

$ perl6 -e 'my Int​:D %a; %a<a> := Str; %a.say;'
{a => (Str)}

...note how the array version fails (correctly) in the binder​:

$ perl6 -e 'my Int​:D @​a; @​a[0] := Str; @​a.say;'
Cannot resolve caller BIND-POS(Array[Int​:D]​: Int, Str); none of these signatures match​:
  (Array​:D $​: Int $pos, Int​:D \bindval, *%_)
  (Array​:D $​: int $pos, Int​:D \bindval, *%_)
  in block <unit> at -e line 1
$ perl6 -e 'my Int​:D @​a; @​a.^find_method("BIND-POS").candidates».signature.say'
((Array​:D $​: Int $pos, Int​:D \bindval, *%_) (Array​:D $​: int $pos, Int​:D \bindval, *%_))
$ perl6 -e 'my Int​:D %a; %a.^find_method("BIND-KEY").candidates».signature.say'
((Hash​:D $​: \key, Mu \bindval, *%_))

...and note in Hash.pm there are comments on the hash class BIND-KEY method saying
it should be a multi, but that blows up setting compilation -- I verified this is still
currently the case. That might be related.

One thing the awesomeized errors do not get right is this, but it is in a half broken,
half NYI state anyway​:

$ perl6 -e 'role A[​::T $t] { has T​:D $.f = 1; }; A[Int].new'
X​::TypeCheck​::Assignment exception produced no message
  in any at gen/moar/m-Metamodel.nqp line 1736
  in block <unit> at -e line 1

(versus current behavior, also wrong as it should not fail, if ever implemented)​:

(03​:31​:08 PM) skids​: m​: role A[​::T $t] { has T​:D $.f = 1; }; A[Int].new
(03​:31​:09 PM) camelia​: rakudo-moar 478671​: OUTPUT«Type check failed in assignment to $!f; expected T​:D but got Int (1)␤ in any at gen/moar/m-Metamodel.nqp line 1736␤ in block <unit> at <tmp> line 1␤␤»

Evidence that it should not be hoped to work​:

$ perl6 -e 'role A[​::T $t] { method f { (T​:D).perl.say } }; A[Int].new.f()'
===SORRY!=== Error while compiling -e
Type too complex to form a definite type
at -e​:1

From @zoffixznet

Fixed in rakudo/rakudo@5500404

Tests added in Raku/roast@fa9a6b0

--
Cheers,
ZZ | https://twitter.com/zoffix

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

From @LLFourn

Tests​:
Raku/roast@fa9a6b0

On Mon, Jul 25, 2016 at 8​:19 AM Brian S. Julin via RT <
perl6-bugs-followup@​perl.org> wrote​:

On Fri Jul 22 15​:34​:49 2016, bri@​abrij.org wrote​:

rakudo PR #​829 should fix this on moarvm. JVM and JS fixes may also
be needed.

I added some error awesomeization to that PR so that now​:

$ perl6 -e 'my Int​:D @​a; @​a[0] = Nil; @​a.say;'
Type check failed in assignment to @​a; expected type Int​:D cannot be
itself (perhaps Nil was assigned to a :D which had no default?)
in block <unit> at -e line 1

$ perl6 -e 'my Int​:D @​a; @​a[0] = Int​:D; @​a.say;'
Type check failed in assignment to @​a; expected type Int​:D cannot be
itself (perhaps Nil was assigned to a :D which had no default?)
in block <unit> at -e line 1

$ perl6 -e 'my Int​:D $a = 42; $a := Nil; $a.say;'
Type check failed in binding; expected Int​:D but got Nil (Nil)
in block <unit> at -e line 1

$ perl6 -e 'my Int​:D $a = 42; $a := Int​:D; $a.say;'
Type check failed in binding; expected type Int​:D cannot be itself
in block <unit> at -e line 1

And though there is an awesomeization for sub return types it does not
work,
because the X is being fed the raw type not the smilied typed.

$ perl6 -e '{ sub f() of Str​:D { Str​:D }; f(); CATCH {
$_.expected.WHAT.say }}'
(Str)
Type check failed for return value; expected Str but got Str​:D (Str​:D)
in sub f at -e line 1
in block <unit> at -e line 1

...that is a separate issue so, not touching it.

Another thing I'm not touching is this​:

$ perl6 -e 'my Int​:D %a; %a<a> := Str; %a.say;'
{a => (Str)}

...note how the array version fails (correctly) in the binder​:

$ perl6 -e 'my Int​:D @​a; @​a[0] := Str; @​a.say;'
Cannot resolve caller BIND-POS(Array[Int​:D]​: Int, Str); none of these
signatures match​:
(Array​:D $​: Int $pos, Int​:D \bindval, *%_)
(Array​:D $​: int $pos, Int​:D \bindval, *%_)
in block <unit> at -e line 1
$ perl6 -e 'my Int​:D @​a;
@​a.^find_method("BIND-POS").candidates».signature.say'
((Array​:D $​: Int $pos, Int​:D \bindval, *%_) (Array​:D $​: int $pos, Int​:D
\bindval, *%_))
$ perl6 -e 'my Int​:D %a;
%a.^find_method("BIND-KEY").candidates».signature.say'
((Hash​:D $​: \key, Mu \bindval, *%_))

...and note in Hash.pm there are comments on the hash class BIND-KEY
method saying
it should be a multi, but that blows up setting compilation -- I verified
this is still
currently the case. That might be related.

One thing the awesomeized errors do not get right is this, but it is in a
half broken,
half NYI state anyway​:

$ perl6 -e 'role A[​::T $t] { has T​:D $.f = 1; }; A[Int].new'
X​::TypeCheck​::Assignment exception produced no message
in any at gen/moar/m-Metamodel.nqp line 1736
in block <unit> at -e line 1

(versus current behavior, also wrong as it should not fail, if ever
implemented)​:

(03​:31​:08 PM) skids​: m​: role A[​::T $t] { has T​:D $.f = 1; }; A[Int].new
(03​:31​:09 PM) camelia​: rakudo-moar 478671​: OUTPUT«Type check failed in
assignment to $!f; expected T​:D but got Int (1)␤ in any at
gen/moar/m-Metamodel.nqp line 1736␤ in block <unit> at <tmp> line 1␤␤»

Evidence that it should not be hoped to work​:

$ perl6 -e 'role A[​::T $t] { method f { (T​:D).perl.say } }; A[Int].new.f()'
===SORRY!=== Error while compiling -e
Type too complex to form a definite type
at -e​:1