"Obviously" the script is flawed, but the real question is did you see it before you ran the script or afterwards:-)

For the record: I did afterwards:-/ Gestalt Principles of Perception?

Before. I took the approach of identifying difference with what I would do. I immediately saw two things: code before functions, and use of exit. I looked if there was more code below exit, and voila!

The fix is to add one word:

#!/usr/bin/perl

use strict;
use warnings;

use Data::Dumper;

_test();
exit;

######## subroutines

BEGIN {                           <-----
   my @values = qw( 1 2 3 );

   sub _test {
      print "Values:\n", Dumper( \@values );
   }
}
[download]

I saw it before I ran the script. But obviously, from how the question was framed, I knew in advance there was something fishy. And the shortness of the code didn't leave many options of what would be wrong.

Would I have seen it if the program was larger, and it wasn't flagged as "there's something going on here"? Probably not.

I didn't see it because I was expecting some closure pitfall and was searching for that to the exclusion of everything else.

That's a classic "gotcha" in my book. That's one reason why I almost always write such code as:

{
    my @values;
    BEGIN { @values = ( 1 .. 3 ); }

    sub _test {
        # ...
    }
}
[download]

or, especially if the code is going into the main script for a mod_perl Apache::Registry page:

BEGIN {
    my @values = ( 1 .. 3 );

    sub _test {
        # ...
    }
}
[download]

Actually, that makes me wonder why I've been using the first form lately. I should switch back to the second form. Thanks for reminding me.

Oh, and to be extra clear, yes, it is easy to get caught by this problem (which is why I have a "practice" for preventing it).