From @masak

<skids> r​: role A { method pub { self!priv }; method !priv () {
"OHAI".say } }; class C does A { }; C.new.pub;
<camelia> rakudo-jvm a77214​: OUTPUT«Nominal type check failed for
parameter 'null'␤␤»
<camelia> ..rakudo-parrot a77214​: OUTPUT«Nominal type check failed for
parameter ''; expected $?CLASS but got C instead␤ in method priv at
/tmp/eL6iqOxDYJ​:1␤ in method pub at /tmp/eL6iqOxDYJ​:1␤ in block at
/tmp/eL6iqOxDYJ​:1␤␤»
<skids> r​: role A { method pub { self!priv }; method !priv ($self​:) {
"OHAI".say } }; class C does A { }; C.new.pub;
<camelia> rakudo-parrot a77214, rakudo-jvm a77214​: OUTPUT«OHAI␤»
<masak> skids​: that... looks like a bug to me. at first blush.
<masak> skids​: looks like 'self' isn't parameterized correctly in
private methods in roles.
<masak> r​: class A { method pub { self!priv }; method !priv { say "OH
HAI" } }; A.new.pub
<camelia> rakudo-parrot a77214, rakudo-jvm a77214​: OUTPUT«OH HAI␤»
<masak> skids​: do you know if that's already submitted?
<skids> masak​: I do not. I was removing old workarounds from a year +
ago and found it.
* masak submits rakudobug
<lue> masak​: wait, how is that a bug? Seems reasonable to me.
<masak> lue​: note, it works in my last eval.
<masak> lue​: but it doesn't work in skid's first.
<masak> lue​: the only difference is that in skid's case, pub and !priv
are in a role.
<lue> masak​: oh, I thought you were discussing skids' last eval.
nevermind then :) [somehow I also missed the $self​: in that eval]
<masak> the second has a $self​:
<masak> the first doesn't.
<masak> the fact that those two differ in their result is also suspect.
<skids> At least the $self is a better workaround now than using a
public method was.

From @usev6

This works on Moar but fails on Parrot and JVM​:

$ perl6-m -e 'role A { method pub { self!priv }; method !priv () { say "OH HAI" } }; class C does A { }; C.new.pub;'
OH HAI

$ perl6-p -e 'role A { method pub { self!priv }; method !priv () { say "OH HAI" } }; class C does A { }; C.new.pub;'
Cannot type check against type variable $?CLASS
  in any type_check at gen/parrot/Metamodel.nqp​:297
  in block at gen/parrot/CORE.setting​:963
  in method BUILDALL at gen/parrot/CORE.setting​:946
  in method bless at gen/parrot/CORE.setting​:935
  in method new at gen/parrot/CORE.setting​:920
  in method new at gen/parrot/CORE.setting​:918
  in method priv at -e​:1
  in method pub at -e​:1
  in block <unit> at -e​:1

$ perl6-j -e 'role A { method pub { self!priv }; method !priv () { say "OH HAI" } }; class C does A { }; C.new.pub;'
Cannot type check against type variable $?CLASS
  in any type_check at gen/jvm/Metamodel.nqp​:297
  in block at gen/jvm/CORE.setting​:949
  in method BUILDALL at gen/jvm/CORE.setting​:932
  in method bless at gen/jvm/CORE.setting​:921
  in method new at gen/jvm/CORE.setting​:905
  in method priv at -e​:1
  in method pub at -e​:1
  in block <unit> at -e​:1

I added a test (skipped for Parrot and JVM) to S14-roles/basic.t with the following commit​: Raku/roast@8942c27ab2

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

From @lizmat

On 17 Oct 2014, at 14​:45, Christian Bartolomaeus via RT <bugs-comment@​bugs6.perl.org> wrote​:

This works on Moar but fails on Parrot and JVM​:

$ perl6-m -e 'role A { method pub { self!priv }; method !priv () { say "OH HAI" } }; class C does A { }; C.new.pub;'
OH HAI

$ perl6-p -e 'role A { method pub { self!priv }; method !priv () { say "OH HAI" } }; class C does A { }; C.new.pub;'
Cannot type check against type variable $?CLASS
in any type_check at gen/parrot/Metamodel.nqp​:297
in block at gen/parrot/CORE.setting​:963
in method BUILDALL at gen/parrot/CORE.setting​:946
in method bless at gen/parrot/CORE.setting​:935
in method new at gen/parrot/CORE.setting​:920
in method new at gen/parrot/CORE.setting​:918
in method priv at -e​:1
in method pub at -e​:1
in block <unit> at -e​:1

$ perl6-j -e 'role A { method pub { self!priv }; method !priv () { say "OH HAI" } }; class C does A { }; C.new.pub;'
Cannot type check against type variable $?CLASS
in any type_check at gen/jvm/Metamodel.nqp​:297
in block at gen/jvm/CORE.setting​:949
in method BUILDALL at gen/jvm/CORE.setting​:932
in method bless at gen/jvm/CORE.setting​:921
in method new at gen/jvm/CORE.setting​:905
in method priv at -e​:1
in method pub at -e​:1
in block <unit> at -e​:1

I added a test (skipped for Parrot and JVM) to S14-roles/basic.t with the following commit​: Raku/roast@8942c27ab2

Related​: $ 6 'role R { method !m(@​a) { say "ok​: @​a[]" }; method x() { self!m((my %).values) } }; class C does R { }; C.x'
No such method 'item' for invocant of type '$?CLASS'
  in method message at src/gen/m-CORE.setting​:13330
  in method m at -e​:1
  in method x at -e​:1
  in block <unit> at -e​:1

This only seems to happen when passing something like .values of a hash.

From @skids

This is a reminder that, when this bug is fixed, there are
workarounds for it to remove in src/core/IO/Local.pm and
(possibly, PR still pending) in src/core/Temporal.pm

In the former look for commented out lines accessig self! things
and use them to replace the workaround code in those methods

In the latter, make the VALID-UNIT method private and adjust
the calls to it.

From @usev6

The test in S14-roles/basic.t passes now on JVM.

Also the code from lizmat++ does not die​:

$ perl6 -e 'role R { method !m(@​a) { say "ok​: @​a[]" }; method x() { self!m((my %).values) } }; class C does R { }; C.x'
ok​:

I haven't yet looked at src/core/IO/Local.pm or src/core/Temporal.pm

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