monitor methods convert die into crazy

[japhb]

For some reason, ad hoc (die "...") exceptions get swallowed and replaced with a 'Cannot invoke this object (REPR: Null; VMNull)' error; here's a golf:

$ cat A.pm6
use OO::Monitors;
unit monitor A;
method B { say "C"; die "D"; "E" }

$ perl6 -I . -e 'need A; say A.new.B'
C
Cannot invoke this object (REPR: Null; VMNull)
  in block  at .../install/share/perl6/site/sources/B7DF632EBAC65169FAE1AF8BFB774E9775D8BA37 (OO::Monitors) line 33
  in block  at .../install/share/perl6/site/sources/B7DF632EBAC65169FAE1AF8BFB774E9775D8BA37 (OO::Monitors) line 30
  in any enter at gen/moar/m-Metamodel.nqp line 4032
  in block <unit> at -e line 1

Given that the current version of OO::Monitors specifically claims to work around "Cannot invoke this object" errors, something is fishy.

[japhb]

OK, it may have nothing to do with OO::Monitors per se, just something OO::Monitors happens to do that triggers a different bug. Attacking the golfing problem from another angle, I reached this arrangement:

$ cat K.pm6
use C;
sub L is export { C::D }

$ cat C.pm6
unit module C;

my %h;

INIT %h<c> = -> $x?, $y? { $x + $y };

our sub D {
    say "A";
    say %h<c>();
    say "B";
}

$ perl6 -I . -e "use K; L"
A
Cannot invoke this object (REPR: Null; VMNull)
  in block  at .../Terminal-Print/C.pm6 (C) line 5
  in sub D at .../Terminal-Print/C.pm6 (C) line 9
  in sub L at .../K.pm6 (K) line 2
  in block <unit> at -e line 1

This turns out to be remarkably finicky about the contents of C.pm6. For example, instead of storing the pointy in a hash entry, let's just bind it to a scalar:

$ cat C.pm6
unit module C;

my $h;

INIT $h := -> $x?, $y? { $x + $y };

our sub D {
    say "A";
    say $h();
    say "B";
}

$ perl6 -I . -e "use K; L"
A
Cannot find method 'Any' on object of type List
  in sub D at .../Terminal-Print/C.pm6 (C) line 9
  in sub L at .../Terminal-Print/K.pm6 (K) line 2
  in block <unit> at -e line 1

OK, instead of binding, we'll assign:

$ cat C.pm6
unit module C;

my $h;

INIT $h = -> $x?, $y? { $x + $y };

our sub D {
    say "A";
    say $h();
    say "B";
}

$ perl6 -I . -e "use K; L"
A
Cannot resolve caller infix:<+>(Mu, Mu); none of these signatures match:
    ($x = 0)
    (\a, \b)
    (Real \a, Real \b)
    (Int:D \a, Int:D \b --> Int:D)
    (int $a, int $b --> int)
    (Num:D \a, Num:D \b)
    (num $a, num $b --> num)
    (Range:D \a, Real:D \b)
    (Real:D \a, Range:D \b)
    (Rational \a, Rational \b)
    (Rational \a, Int \b)
    (Int \a, Rational \b)
    (Complex:D \a, Complex:D \b --> Complex:D)
    (Complex:D \a, Real \b --> Complex:D)
    (Real \a, Complex:D \b --> Complex:D)
    (Instant:D $a, Real:D $b)
    (Real:D $a, Instant:D $b)
    (Instant:D $a, Duration:D $b)
    (Duration:D $a, Instant:D $b)
    (Duration:D $a, Real $b)
    (Real $a, Duration:D $b)
    (Duration:D $a, Duration:D $b)
    (Date:D $d, Int:D $x)
    (Int:D $x, Date:D $d)
  in block  at .../Terminal-Print/C.pm6 (C) line 5
  in sub D at .../Terminal-Print/C.pm6 (C) line 9
  in sub L at .../Terminal-Print/K.pm6 (K) line 2
  in block <unit> at -e line 1

OK, let's supply sane arguments:

$ cat C.pm6
unit module C;

my $h;

INIT $h = -> $x?, $y? { $x + $y };

our sub D {
    say "A";
    say $h(1,2);
    say "B";
}

$ perl6 -I . -e "use K; L"
A
3
B

Well, that works! But wait, what about if I use a hash again, just with proper arguments this time?

$ cat C.pm6
unit module C;

my %h;

INIT %h<c> = -> $x?, $y? { $x + $y };

our sub D {
    say "A";
    say %h<c>(1,2);
    say "B";
}

$ perl6 -I . -e "use K; L"
A
Cannot invoke this object (REPR: Null; VMNull)
  in block  at .../Terminal-Print/C.pm6 (C) line 5
  in sub D at .../Terminal-Print/C.pm6 (C) line 9
  in sub L at .../Terminal-Print/K.pm6 (K) line 2
  in block <unit> at -e line 1

Boom! What if I get rid of INIT?

$ cat C.pm6
unit module C;

my %h;

%h<c> = -> $x?, $y? { $x + $y };

our sub D {
    say "A";
    say %h<c>(1,2);
    say "B";
}

$ perl6 -I . -e "use K; L"
A
3
B

And now it works again. WAT.

[jonathanstowe]

Is this a slight variation on https://rt.perl.org/Ticket/Display.html?id=127858 which seems to be that some Code objects don't survive pre-compilation?

[japhb]

Yeah, I was thinking something similar and then saw your comment, jonathanstowe. Just to confirm, I went back to the second-to-last version above, the one with INIT that fails, and just add no precompilation; at the top:

$ cat C.pm6
no precompilation;

unit module C;

my %h;

INIT %h<c> = -> $x?, $y? { $x + $y };

our sub D {
    say "A";
    say %h<c>(1,2);
    say "B";
}

$ perl6 -I . -e "use K; L"
A
3
B

Sure enough, it works. So there are at least two bugs here: A confirmation of RT #127858 with a few more variants, and that the workaround that OO::Monitors uses doesn't quite work (see the original issue description).

[jnthn]

@japhb So I'm a tad confused. I turned the original issue reported here into what I'm pretty sure is a representative pre-comp test case. The test failed. Then I pulled in the changes from my proposed maybe-fix-exception-stuff branch. That made the test pass. So, that's in the commit I linked.

However, I thought you'd mentioned that the branch did not help? So, I tried your original example:

jnthn@lviv:~/dev/oo-monitors$ cat A.pm6 
use OO::Monitors;
unit monitor A;
method B { say "C"; die "D"; "E" }
jnthn@lviv:~/dev/oo-monitors$ perl6 -I . -e 'need A; say A.new.B'
C
D
  in method B at /home/jnthn/dev/oo-monitors/A.pm6 (A) line 3
  in any call_with_capture at gen/moar/Metamodel.nqp line 3901
  in block  at /home/jnthn/dev/oo-monitors/lib/OO/Monitors.pm6 (OO::Monitors) line 31
  in any enter at gen/moar/Metamodel.nqp line 3999
  in block <unit> at -e line 1

This case seems fixed also (so the test was covering the right thing). Yes, there is a Rakudo bug that also wants looking at, but in all those cases OO::Monitors has been golfed out of the issue.