sybperl-l Archive

From: Michael Peppler <mpeppler at MBAY dot NET>
Subject: Sybperl memory leak
Date: Feb 11 1999 6:43PM

>>>>> "Steve" == Wechsler, Steven M  writes:

Steve> I ran this program on NT 4.0 running ActivePerl 5.00502 and the
Steve> latest version of Sybperl available from the Activestate web
Steve> site.

Steve> Each iteration of the loop caused 64K more memory to be
Steve> allocated.  Michael, your web site mentions a 300 byte memory
Steve> leak; any idea how 300 bytes becomes 64K?

Actually I re-did the test here, and I get a 2.9k leak per iteration
(not good, but better than 64k). This is on Solaris 7, OC 11.1.1. I
don't know how much of that is in the Sybase code, and how much is in
the perl code.

Right now I don't really have the time nor the tools to chase this
down. If someone with purify or some similar tools wants to give it a
go I'd be more than happy to hear from you.

In the meantime (at least on Unix systems) your script could be
re-written like this to avoid excessive bloating:

use Sybase::DBlib;

for(my $count = 1;; $count++) {
    sleep(10);
    my $dbh = new Sybase::DBlib $user, $pwd, $server;
    warn "$server is down or unreachable\n" unless $dbh;

    if(($count % 100) == 0) {
	exec "$0 @ARGV";
    }
}
__END__

By using the "my $dbh" in the loop the dbclose() call will be done
automatically.

By putting the sleep() first the connection doesn't stay open any
longer than necessary.

The exec() call re-exec's the script every 100 calls, which should
keep the total memory usage in check (I don't know if exec() works on
Win32 platforms, though).

Michael
-- 
Michael Peppler         -||-  Data Migrations Inc.
mpeppler@mbay.net       -||-  http://www.mbay.net/~mpeppler
Int. Sybase User Group  -||-  http://www.isug.com
Sybase on Linux mailing list: ase-linux-list@isug.com