make in regex on uncomposed type results in Nil

[p6rt]

Migrated from rt.perl.org#132512 (status was 'open')

Searchable as RT132512$

[p6rt]

From @LLFourn

Just got around to investigating​: https://rt-archive.perl.org/perl6/Ticket/Dis
play.html?id=132085

It turns out to be a regression.

my $new_type := Metamodel​::ClassHOW.new_type(​:name<Foo>);
my $r = / . { make $new_type } /;
my $m = "a" ~~ $r;
note $m.ast; #-> Nil

making an uncomposed type somehow results in Nil now. It used to work. In
my compiler I can't really get around needing this to work.

LL

[p6rt]

From @AlexDaniel

What do you mean exactly by “used to work”? Here's the output on all 6c releases​: https://gist.github.com/efee7716c35d36c6f793465c2f0b6035

Which behavior is right? Or what's would be the right snippet to reproduce it?

On 2017-11-27 18​:48​:07, lloyd.fourn@​gmail.com wrote​:

Just got around to investigating​: https://rt-archive.perl.org/perl6/Ticket/Dis
play.html?id=132085

It turns out to be a regression.

my $new_type := Metamodel​::ClassHOW.new_type(​:name<Foo>);
my $r = / . { make $new_type } /;
my $m = "a" ~~ $r;
note $m.ast; #-> Nil

making an uncomposed type somehow results in Nil now. It used to work. In
my compiler I can't really get around needing this to work.

LL

[p6rt]

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

[p6rt]

From @LLFourn

Good point. Here "No such method 'gist' for invocant of type 'Foo'
  in block <unit> at /tmp/aR11azfzlJ line 1" is the right one.

This will give True/False indicating correct/incorrect​:

my $new_type := Metamodel​::ClassHOW.new_type(​:name<Foo>);
my $r = / . { $/.make($new_type) } /;
my $m = "a" ~~ $r;
note $m.ast.^name eq "Foo";

Thanks for bisecting magic!

On Tue, Nov 28, 2017 at 1​:54 PM Aleks-Daniel Jakimenko-Aleksejev via RT <
perl6-bugs-followup@​perl.org> wrote​:

What do you mean exactly by “used to work”? Here's the output on all 6c
releases​: https://gist.github.com/efee7716c35d36c6f793465c2f0b6035

Which behavior is right? Or what's would be the right snippet to reproduce
it?

On 2017-11-27 18​:48​:07, lloyd.fourn@​gmail.com wrote​:

Just got around to investigating​: https://rt-archive.perl.org/perl6/Ticket/Dis
play.html?id=132085

It turns out to be a regression.

my $new_type := Metamodel​::ClassHOW.new_type(​:name<Foo>);
my $r = / . { make $new_type } /;
my $m = "a" ~~ $r;
note $m.ast; #-> Nil

making an uncomposed type somehow results in Nil now. It used to work. In
my compiler I can't really get around needing this to work.

LL

[p6rt]

From @AlexDaniel

OK, the change from True to False happened here​: (2017-08-21) rakudo/rakudo@5db5b1d

I'm not entirely convinced that the current behavior is incorrect, but then again I'm sleep deprived. Maybe someone else will have a better idea.

On 2017-11-27 19​:00​:45, lloyd.fourn@​gmail.com wrote​:

Good point. Here "No such method 'gist' for invocant of type 'Foo'
in block <unit> at /tmp/aR11azfzlJ line 1" is the right one.

This will give True/False indicating correct/incorrect​:

my $new_type := Metamodel​::ClassHOW.new_type(​:name<Foo>);
my $r = / . { $/.make($new_type) } /;
my $m = "a" ~~ $r;
note $m.ast.^name eq "Foo";

Thanks for bisecting magic!

On Tue, Nov 28, 2017 at 1​:54 PM Aleks-Daniel Jakimenko-Aleksejev via
RT <
perl6-bugs-followup@​perl.org> wrote​:

What do you mean exactly by “used to work”? Here's the output on all
6c
releases​: https://gist.github.com/efee7716c35d36c6f793465c2f0b6035

Which behavior is right? Or what's would be the right snippet to
reproduce
it?

On 2017-11-27 18​:48​:07, lloyd.fourn@​gmail.com wrote​:

Just got around to investigating​: https://rt-archive.perl.org/perl6/Ticket/Dis
play.html?id=132085

It turns out to be a regression.

my $new_type := Metamodel​::ClassHOW.new_type(​:name<Foo>);
my $r = / . { make $new_type } /;
my $m = "a" ~~ $r;
note $m.ast; #-> Nil

making an uncomposed type somehow results in Nil now. It used to
work. In
my compiler I can't really get around needing this to work.

LL

[p6rt]

From @LLFourn

Changing uncomposed type objects to Nil for the specific case of .ast/.made
is definitely incorrect IMO.

It happens for the exact reason lizmat said in her commit​: "Wish there was
a better way to test for NQPMu
though, as this will prevent type objects being properly propagated"

Although it looks like this has been fixed for composed type objects in a
following patch. It now looks like​:

method ast() { nqp​::if(nqp​::istype($!made, Mu),$!made,Nil) }

But since uncomposed type objects don't inherit from Mu we have the bug.

On Tue, Nov 28, 2017 at 3​:26 PM Aleks-Daniel Jakimenko-Aleksejev via RT <
perl6-bugs-followup@​perl.org> wrote​:

OK, the change from True to False happened here​: (2017-08-21)

rakudo/rakudo@5db5b1d

I'm not entirely convinced that the current behavior is incorrect, but then
again I'm sleep deprived. Maybe someone else will have a better idea.

On 2017-11-27 19​:00​:45, lloyd.fourn@​gmail.com wrote​:

Good point. Here "No such method 'gist' for invocant of type 'Foo'
in block <unit> at /tmp/aR11azfzlJ line 1" is the right one.

This will give True/False indicating correct/incorrect​:

my $new_type := Metamodel​::ClassHOW.new_type(​:name<Foo>);
my $r = / . { $/.make($new_type) } /;
my $m = "a" ~~ $r;
note $m.ast.^name eq "Foo";

Thanks for bisecting magic!

On Tue, Nov 28, 2017 at 1​:54 PM Aleks-Daniel Jakimenko-Aleksejev via
RT <
perl6-bugs-followup@​perl.org> wrote​:

What do you mean exactly by “used to work”? Here's the output on all
6c
releases​: https://gist.github.com/efee7716c35d36c6f793465c2f0b6035

Which behavior is right? Or what's would be the right snippet to
reproduce
it?

On 2017-11-27 18​:48​:07, lloyd.fourn@​gmail.com wrote​:

Just got around to investigating​: https://rt-archive.perl.org/perl6/Ticket/Dis
play.html?id=132085

It turns out to be a regression.

my $new_type := Metamodel​::ClassHOW.new_type(​:name<Foo>);
my $r = / . { make $new_type } /;
my $m = "a" ~~ $r;
note $m.ast; #-> Nil

making an uncomposed type somehow results in Nil now. It used to
work. In
my compiler I can't really get around needing this to work.

LL