On Tue, Jan 12, 2010 at 9:37 AM, Nele Kosog <k...@sevencs.com> wrote:
> Since the Digest method is passed as string, I think Data::Serializer uses
> "SHA-256" (that's the default and I don't pass a different parameter to
> Data::Serializer).
> ...
> No, it doesn't work with -M Digest::SHA.
Referring to the side note in my last post (sorry didn't realize
the consequences): par.pl doing
eval { $Digest::SHA::VERSION };
totally screws Digest's attempt to autoload Digest::SHA.
Snippet from Digest.pm, $class holds the name of
the actual digester module that we want to use, in your
case "Digest::SHA":
no strict 'refs';
unless (exists ${"$class\::"}{"VERSION"}) {
eval "require $class";
if ($@) {
$err ||= $@;
next;
}
}
return $class->new(@args, @_);
Since par.pl has already been run (from prelude code in the packed
executable that is run before your actual script), the above test
has the side effect of bringing the global variable $Digest::SHA::VERSION
"to life" (as if you had said "package Digest::SHA; our $VERSION; ...").
That variable actually is stored in the strange hash
%{"Digest::SHA::"} under the key "VERSION". Hence the exists()
returns true and Digest skips "require Digest::SHA".
So while pp option -M generally works for modules whose usage
goes undetected by both static dynamic (option -x) analysis, in your
case it causes pp to pack the module, but Digest simply fails to load it.
The simplest workaround is for you to explicitly
use Digest::SHA;
somewhere in your script (and then you can forget about -M).
Cheers, Roderich
