Front page | perl.perl5.porters |
Postings from July 2013
On 28 July 2013 09:19, Father Chrysostomos via RT
<perlbug-followup@perl.org> wrote:
> On Sat Jul 27 17:05:39 2013, rjbs wrote:
>> I received a user report along with an excellent reproducer, here
>> converted to a test:
>>
>> use PadWalker 'peek_my';
>> use Test::More;
>>
>> my ($var1, $var2, $var3);
>>
>> our $want = { map {; $_ => \undef } qw($var1 $var2 $var3) };
>>
>> sub { is_deeply( peek_my(0), $want, 'anon A' ); }-
>> >();
>
> That one is not a closure. It passes (serendipitously).
>
>> sub { is_deeply( peek_my(0), $want, 'anon B' ); $var2 }-
>> >();
>
> That one is. It fails.
>
> It was probably a0d2bbd5c470 that did it. See #89544.
>
> PadWalker simply follows CvOUTSIDE pointers. The problem with a closure
> clone having an outside pointer is that it has to be reference-counted,
> easily resulting in circularities.
>
> We must retain the CvOUTSIDE pointer for string eval to be able to look
> up variables in the parent. But in the absence of string eval or s///ee
> we try to avoid memory leaks. Also, try running it under the debugger.
> It passes.
>
> I do not consider this a bug, because the closure does not close over
> $var1 or $var3. Those variables may already have been freed by the time
> you call peek_my.
While I agree with your analysis I thought that PADMY vars werent
freed until global destruction.
Yves
--
perl -Mre=debug -e "/just|another|perl|hacker/"
Thread Previous
|
Thread Next