From @masak

<masak> rakudo​: sub foo(@​a) { @​a[1] = "NOES" }; my @​a = <OH HAI>;
foo(@​a); say ~@​a
<p6eval> rakudo 1eef08​: OUTPUT«OH NOES␤»
<masak> is this supposed to be allowed?
<pmichaud> masak​: I haven't seen a clear answer to that question.
<masak> pmichaud​: ISTR there was some wording in the spec about the
readonlyness stretching 'one level down' into arrays/hashes, but I
can't find it now, despite trying.
<pmichaud> masak​: From an implementation perspective, I see nothing
but pain in trying to make that happen.
<masak> pmichaud​: anyway, this already fails​:
<masak> rakudo​: sub foo(@​a) { @​a[0, 1] = <OH NOES> }; my @​a = 0, 1; foo(@​a)
<p6eval> rakudo 1eef08​: OUTPUT«Cannot assign to readonly value [...]
<masak> so something is inconsistent, at least.
* masak submits rakudobug

From @coke

On Tue May 04 14​:13​:17 2010, masak wrote​:

<masak> rakudo​: sub foo(@​a) { @​a[1] = "NOES" }; my @​a = <OH HAI>;
foo(@​a); say ~@​a
<p6eval> rakudo 1eef08​: OUTPUT«OH NOES␤»
<masak> is this supposed to be allowed?
<pmichaud> masak​: I haven't seen a clear answer to that question.
<masak> pmichaud​: ISTR there was some wording in the spec about the
readonlyness stretching 'one level down' into arrays/hashes, but I
can't find it now, despite trying.
<pmichaud> masak​: From an implementation perspective, I see nothing
but pain in trying to make that happen.
<masak> pmichaud​: anyway, this already fails​:
<masak> rakudo​: sub foo(@​a) { @​a[0, 1] = <OH NOES> }; my @​a = 0, 1; foo(@​a)
<p6eval> rakudo 1eef08​: OUTPUT«Cannot assign to readonly value [...]
<masak> so something is inconsistent, at least.
* masak submits rakudobug

Behavior changed​:

sub foo(@​a) { @​a[1] = "NOES" }; my @​a = <OH HAI>;
OH HAI

sub foo(@​a) { @​a[0, 1] = <OH NOES> }; my @​a = 0, 1; foo(@​a)
OH NOES

--
Will "Coke" Coleda

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

From @coke

On Sat Oct 20 18​:12​:38 2012, coke wrote​:

On Tue May 04 14​:13​:17 2010, masak wrote​:

<masak> rakudo​: sub foo(@​a) { @​a[1] = "NOES" }; my @​a = <OH HAI>;
foo(@​a); say ~@​a
<p6eval> rakudo 1eef08​: OUTPUT«OH NOES␤»
<masak> is this supposed to be allowed?
<pmichaud> masak​: I haven't seen a clear answer to that question.
<masak> pmichaud​: ISTR there was some wording in the spec about the
readonlyness stretching 'one level down' into arrays/hashes, but I
can't find it now, despite trying.
<pmichaud> masak​: From an implementation perspective, I see nothing
but pain in trying to make that happen.
<masak> pmichaud​: anyway, this already fails​:
<masak> rakudo​: sub foo(@​a) { @​a[0, 1] = <OH NOES> }; my @​a = 0, 1;
foo(@​a)
<p6eval> rakudo 1eef08​: OUTPUT«Cannot assign to readonly value [...]
<masak> so something is inconsistent, at least.
* masak submits rakudobug

Behavior changed​:

sub foo(@​a) { @​a[1] = "NOES" }; my @​a = <OH HAI>;
OH HAI

sub foo(@​a) { @​a[0, 1] = <OH NOES> }; my @​a = 0, 1; foo(@​a)
OH NOES

These are now the same (both in nom and glr, even)

12​:10 < [Coke]> m​: sub foo(@​a) { @​a[1] = "NOES" }; my @​a = <OH HAI>; foo(@​a); say ~@​a
12​:10 <+camelia> rakudo-moar 47ddca​: OUTPUT«OH NOES␤»
12​:10 < GLRelia> rakudo-moar 345953​: OUTPUT«OH NOES␤»
12​:11 < [Coke]> m​: sub foo(@​a) { @​a[0, 1] = <OH NOES> }; my @​a = 0, 1; foo(@​a); say ~@​a;
12​:11 <+camelia> rakudo-moar 47ddca​: OUTPUT«OH NOES␤»
12​:11 < GLRelia> rakudo-moar 345953​: OUTPUT«OH NOES␤»

Closable with tests?
--
Will "Coke" Coleda

From @masak

<masak> [Coke]​: no.
<masak> [Coke]​: I think you misunderstood the goal of the ticket.
<masak> [Coke]​: the spec said (and still *says*, I think) that the *contents* of the array parameter are readonly. that's what the "one level down" thing refers to.
<masak> [Coke]​: However, the default forces any container argument to also
<masak> be treated as an immutable value. This extends down only one level;
<masak> an immutable container may always return an element that is mutable if
<masak> it so chooses.
<masak> S06​:718
<synbot6> Link​: http://design.perl6.org/S06.html#line_718
* masak adds this to the ticket

From @jnthn

On Thu Aug 27 09​:22​:02 2015, masak wrote​:

<masak> [Coke]​: no.
<masak> [Coke]​: I think you misunderstood the goal of the ticket.
<masak> [Coke]​: the spec said (and still *says*, I think) that the
*contents* of the array parameter are readonly. that's what the "one
level down" thing refers to.
<masak> [Coke]​: However, the default forces any container argument to
also
<masak> be treated as an immutable value. This extends down only one
level;
<masak> an immutable container may always return an element that is
mutable if
<masak> it so chooses.
<masak> S06​:718
<synbot6> Link​: http://design.perl6.org/S06.html#line_718
* masak adds this to the ticket

Resolution​: not this Christmas. Short discussion​:

http://irclog.perlgeek.de/perl6/2015-12-16#i_11729677

In my view, the best hope of doing this is probably through some static analysis, which we can introduce conditional on language version, so it'd be opt in by language version and we'd be able to not break existing code that didn't opt in to using that version.

For the meantime, the ticket survives, but is off the xmas list.

This still a problem, far as I can tell.

Given jnthn++'s verdict above, I think this issue is now a feature request, not a bug report.