in reply to Re^3: Parallel::ForkManager run_on_finish exit code handler
in thread Parallel::ForkManager run_on_finish exit code handler

ikegami - Thanks. The following code works.

for my $host (@hosts) { $pm->start($host) and next; exec(@some_command); print(STDERR "exec failed: $!\n"); _exit($!); }

The database connection is now persistent in run_on_finish(), and I am able to insert the return code into the db from the callback routine.

Thinking of the Chinese proverb.. give a man a fish.. teach a man to fish.. Rather than me just blindly accepting the code improvement, can you explain how run_on_finish() gets the return code without having $return_code passed in as either $pm->finish($return_code) or _exit($return_code) ?

Also, why is 'exec' preferred over 'system + _exit' ? I thought exec executes a system command and never returns.. I am surprised that this works actually.

Replies are listed 'Best First'.
Re^5: Parallel::ForkManager run_on_finish exit code handler
by ikegami (Patriarch) on Sep 09, 2009 at 00:25 UTC

    Your premise that the child process doesn't call exit is wrong. Either it exited properly by calling exit with an exit code (setting $?'s high byte) or it was terminated by a signal (setting $?'s low byte). There's no other way for it to exit.

    I thought exec executes a system command and never returns..

    Same goes for the call to exit it's replacing.

    Also, why is 'exec' preferred over 'system + _exit' ?

    You're creating a process whose sole purpose is to do nothing but create a process, and you're doing that for as many hosts as you have in your array. It's wasteful! It also requires more code and introduces problems (refer to the bug I mentioned it fixed).