From @masak

Rakudo r32629 sometimes dies when assigning a readonly variable to
itself, and sometimes not.

$ ./perl6 -e 'for <a b c> -> $foo { $foo = $foo; say $foo }' # dies, good
Cannot assign to readonly variable.
[...]

$ ./perl6 -e 'for split(/<sp>/, "a b c") -> $foo { $foo = $foo; say
$foo }' # baaaahd!
a
b
c

From @pmichaud

On Thu, Nov 13, 2008 at 03​:21​:45PM -0800, Carl Mäsak wrote​:

Rakudo r32629 sometimes dies when assigning a readonly variable to
itself, and sometimes not.

$ ./perl6 -e 'for <a b c> -> $foo { $foo = $foo; say $foo }' # dies, good
Cannot assign to readonly variable.
[...]

$ ./perl6 -e 'for split(/<sp>/, "a b c") -> $foo { $foo = $foo; say
$foo }' # baaaahd!
a
b
c

There's a bit more going on behind the scenes here -- the "$foo = $foo"
line in the second case is also promoting Parrot String values into
Rakudo Str objects. I suspect that promotion is causing the
readonly property on $foo to be lost.

My guess is that we want to adjust Parrot's C<copy> opcode so that
it doesn't modify the destination PMC's properties.

Thanks,

Pm

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

From @jnthn

On Fri Nov 14 05​:58​:21 2008, pmichaud wrote​:

On Thu, Nov 13, 2008 at 03​:21​:45PM -0800, Carl Mäsak wrote​:

Rakudo r32629 sometimes dies when assigning a readonly variable to
itself, and sometimes not.

$ ./perl6 -e 'for <a b c> -> $foo { $foo = $foo; say $foo }' # dies,
good
Cannot assign to readonly variable.
[...]

$ ./perl6 -e 'for split(/<sp>/, "a b c") -> $foo { $foo = $foo; say
$foo }' # baaaahd!
a
b
c

There's a bit more going on behind the scenes here -- the "$foo = $foo"
line in the second case is also promoting Parrot String values into
Rakudo Str objects. I suspect that promotion is causing the
readonly property on $foo to be lost.

My guess is that we want to adjust Parrot's C<copy> opcode so that
it doesn't modify the destination PMC's properties.

I've fixed it in a patch that's about to go in as soon as it smokes;
please can somebody make sure that a regression test goes in for this.

Thanks,

Jonathan

From @moritz

jnthn@​jnthn.net via RT wrote​:

I've fixed it in a patch that's about to go in as soon as it smokes;
please can somebody make sure that a regression test goes in for this.

test added in (pugs) r23032.

I hereby encourage all Rakudo hackers to give tickets to me if they have
fixed it, but there are no tests yet (with an appropriate comment, of
course). I can't promise to write tests for everything, but the chances
are much higher that way.

Cheers,
Moritz

--
Moritz Lenz
http://perlgeek.de/ | http://perl-6.de/ | http://sudokugarden.de/

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

From @pmichaud

On Mon, Nov 17, 2008 at 10​:15​:09PM +0100, Moritz Lenz wrote​:

I hereby encourage all Rakudo hackers to give tickets to me if they have
fixed it, but there are no tests yet (with an appropriate comment, of
course). I can't promise to write tests for everything, but the chances
are much higher that way.

I go further and make this standard policy​: We don't close a ticket
until there's a test. For now we'll assign "fixed but no test yet"
tickets to moritz.

It would be nice if the ticket also mentioned where the appropriate
test has been placed.

Pm