From @masak

<bbkr_> r​: sub circumfix​:<"[" "]"> ($a) { };
<p6eval> rakudo c82d10​: OUTPUT«Null PMC access in find_method('clone') [...]
<jnthn> ooh.
* masak submits rakudobug
<jnthn> Odd since I'm sure we have passing spectests covering circumfixes...
<masak> r​: sub circumfix​:<" "> ($a) { }
<p6eval> rakudo c82d10​: OUTPUT«Null PMC access in find_method('clone') [...]
<masak> r​: sub circumfix​:<" q> ($a) { }
<p6eval> rakudo c82d10​: OUTPUT«Null PMC access in find_method('clone') [...]
<masak> r​: sub circumfix​:<w "> ($a) { }
<p6eval> rakudo c82d10​: OUTPUT«Null PMC access in find_method('clone') [...]
<masak> so, it doesn't like those double quotes.
<jnthn> r​: sub circumfix​:<w '> ($a) { }
<p6eval> rakudo c82d10​: ( no output )
<jnthn> wtf
<masak> of course single quotes are fine :P
<jnthn> That makes even less sense...

From @usev6

Running the command manually everything is fine (on Parrot, Moar and JVM), but running it as a test resurrects the Null PMC access​:

$ perl6-p -e 'sub circumfix​:<w "> ($a) { }; say q[alive]'
alive

$ PERL6LIB=lib perl6-p -e 'use Test; lives_ok { sub circumfix​:<w "> ($a) { }; }, "running under Test.pm"'
not ok 1 - running under Test.pm

# Failed test 'running under Test.pm'
# at -e line 1
# Null PMC access in find_method('clone')

Running it in REPL is also fine​:

sub circumfix​:<w "> ($a) { }; say "alive"
alive

The nice thing is, that I found out about the problem under Test.pm while adding a test to S06-operator-overloading/sub.t with the following commit​: Raku/roast@18a47abfb3

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

From @FROGGS

Fixed in parrot in commits up to​: parrot/parrot@7bc5452
Tests​: Raku/roast@a8a495c

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