On onsdag, apr 2, 2003, at 16:56 Europe/Stockholm, Bradley W. Langhorst wrote:


On Wed, 2003-04-02 at 08:47, Arthur Bergman wrote:
This has been known to be a bit wonky on Win32, but not under unix
(which I assume you are using because you like forks.pm)

that's correct - i'm using linux 2.4.20


A backtrace would be nice,,


gdb perl
run ./agent.pl
bt

zoiks - that is a long back trace... you may be wondering why so many calls sv_grp and mg_dup...

Im sharing 4 variables between threads
1) fairly complex multi-d hash containing arrays
2) straighforward 2-d hash
3) an array of hashes
4) a hash of arrays


Because the cloning is recursive.


It is not helping us much except pointing out it is in new.


use IO::Socket; use Carp; #could do without this... use threads; use threads::shared; #use forks; #use forks::shared; use Fcntl qw(:DEFAULT :flock); #for locking the log file use Data::Dumper; #for debugging - don't really need this.



my guess is that there is some problem cloning the IO::Socket, could you please try again to get a backtrace with a perl recompiled with debugging


to get one -Doptimize='-g' -Dusethreads should be enough!

Arthur



Reply via email to