PAR: Variable %x will not stay shared

[QM]

I think I know the answer, but I haven't found confirmation...

I have a closure with 3 subs in it, something like this:

use strict;
use warnings;
{ # closure
  my %foo;

  sub collect_it
  { my ($key, $value) = @_;
    $foo{$key} = $value;
  }

  sub print_it
  { for my $key ( keys %foo )
    { print "\$foo{$key} = $foo{$key}\n";
    }
    clear_it();
  }

  sub clear_it
  { %foo = (); }
} # end closure

The script compiles and runs with no warnings or errors, and appears to "do the right thing".

However, some of my users want a PAR executable. When making the PAR executable (pp -z 9 -c script.pl), I get errors for all of the enclosed lexicals:

Variable "%foo" will not stay shared at 60oehl line 123.

...where line "123" is a line in "script.pl" that references %foo.

From my reading I see that this error comes up in Perl when using nested named subs, which isn't the case here.

The only thing I can think of is that PAR is somehow wrapping my whole script, which then triggers the sharing warning message under "-c".

I've tried to reproduce this with a smaller script, but had no luck yet. (The original script is 1000+ lines.) I'm also wondering if the "\$foo{$key}" confuses the compiler into thinking that's a reference?

Any hints?

[Joost]

Apparently PAR wraps your script in a subroutine. I'm not sure if it stores it like that, but if it does, don't ignore the warning - this can be a potential problem.

see this explanation in the mod_perl reference. My preferred solution is to make the base script as short as possible and move all interesting code to a seperate module.

[QM]

It seems to be more complicated than just "PAR wraps your script in a subroutine, and that's the source of the problem." If that were the case, then any program that uses the cache/closure idea should cause the warnings.

The toy script in the OP generates the warnings, but another script where I did something similar did not.

So I'll have to work up an example along the lines of the toy script, but doesn't generate the warnings, and show the minimum tweak that does generate the warnings. (Or some kind soul can chime in with a satisfying explanation.)

[sgifford]

Does:

no warnings 'closure';

solve your problem?

The only thing I can think of is that PAR is somehow wrapping my whole script, which then triggers the sharing warning message under "-c".

I'm not intimately familiar with PAR, but that seems fairly likely.

I've tried to reproduce this with a smaller script, but had no luck yet.

So does the short script you give above reproduce the problem, or no?

[QM]

I've tried to reproduce the problem with a smaller script, but no warnings/errors are generated.

I need to look into some of the modules I'm using, to see if there's any effect there. The toy script doesn't use them, but also doesn't have the problem.

BTW: Yes, no warnings 'closure'; fixes it.

[PodMaster]

Any hints?

Ignore it :)