Andy Armstrong andy at
Wed Mar 1 12:53:04 GMT 2006

On 1 Mar 2006, at 12:30, Aaron Trevena wrote:
> Usual practice is to report any clear bugs on rt, possibly mention
> your feelings in cpan rantings about what you don't like (i.e. the
> stuff that isn't very helpful in a bug report), and send a long
> helpful email to the author - then wait a little more than 48 hours
> for a response.

So it's rude just to rewrite it then? :)

At the moment it walks all the SV arenas at the end of any statement  
in which a new scalar has been created (although actually it misses  
the case in which an equal number of scalars are created and  
destroyed in the same statement) and does some fairly expensive hash  
operations on each one. For example running this code:


use strict;
use warnings;
use Devel::LeakTrace;

$| = 1;

my $x = 1;
my $y = 2;
my $z = \$x;
$x = \$z;
my $l = \($z, $y, $z);

results in 41,334 hash puts and 37,752 hash gets. Add a small array  
and you're suddenly into millions of hash operations for even quite a  
trivial script. So that's more or less all it does at the moment and  
just about all of it needs to change.

I guess I'll finish my version and then discuss it with Richard.

> Uploading a forked version to cpan less than a week after contacting
> the author is a bit of an over-reaction.

Sure. I wasn't intending to do that.

Andy Armstrong,

More information about the mailing list