Launch and control background processes
use Proc::Simple;
$myproc = Proc::Simple->new(); # Create a new process object
$myproc->start("shell-command-line"); # Launch an external program
$myproc->start("command", # Launch an external program
"param", ...); # with parameters
$myproc->start(sub { ... }); # Launch a perl subroutine
$myproc->start(\&subroutine); # Launch a perl subroutine
$myproc->start(\&subroutine, # Launch a perl subroutine
$param, ...); # with parameters
$running = $myproc->poll(); # Poll Running Process
$exit_status = $myproc->wait(); # Wait until process is done
$proc->kill_on_destroy(1); # Set kill on destroy
$proc->signal_on_destroy("KILL"); # Specify signal to be sent
# on destroy
$myproc->kill(); # Kill Process (SIGTERM)
$myproc->kill("SIGUSR1"); # Send specified signal
$myproc->exit_status(); # Return exit status of process
Proc::Simple::debug($level); # Turn debug on
The Proc::Simple package provides objects mimicing real-life processes from a user's point of view. A new process object is created by
$myproc = Proc::Simple->new();
Either external programs or perl subroutines can be launched and controlled as processes in the background.
A 10-second sleep process, for example, can be launched as an external program as in
$myproc->start("/bin/sleep 10"); # or $myproc->start("/bin/sleep", "10");
or as a perl subroutine, as in
sub mysleep { sleep(shift); } # Define mysleep() $myproc->start(\&mysleep, 10); # Launch it.
or even as
$myproc->start(sub { sleep(10); });
The start Method returns immediately after starting the specified process in background, i.e. there's no blocking. It returns 1 if the process has been launched successfully and 0 if not.
The poll method checks if the process is still running
$running = $myproc->poll();
and returns 1 if it is, 0 if it's not. Finally,
$myproc->kill();
terminates the process by sending it the \s-1SIGTERM\s0 signal. As an option, another signal can be specified.
$myproc->kill("SIGUSR1");
sends the \s-1SIGUSR1\s0 signal to the running process. kill returns 1 if it succeeds in sending the signal, 0 if it doesn't.
The methods are discussed in more detail in the next section.
A destructor is provided so that a signal can be sent to the forked processes automatically should the process object be destroyed or if the process exits. By default this behaviour is turned off (see the kill_on_destroy and signal_on_destroy methods).
The following methods are available:
Create a new instance of this class by writing $proc = new Proc::Simple; or $proc = Proc::Simple->new(); It takes no arguments.
Launches a new process. The \*(C`start()\*(C' method can be used to launch both external programs (like \*(C`/bin/echo\*(C') or one of your self-defined subroutines (like \*(C`foo()\*(C') in a new process. For an external program to be started, call $status = $proc->start("program-name"); If you want to pass a couple of parameters to the launched program, there's two options: You can either pass them in one argument like in $status = $proc->start("/bin/echo hello world"); or in several arguments like in $status = $proc->start("/bin/echo", "hello", "world"); Just as in Perl's function \*(C`system()\*(C', there's a big difference between the two methods: If you provide one argument containing a blank-separated command line, your shell is going to process any meta-characters (if you choose to use some) before the process is actually launched: $status = $proc->start("/bin/ls -l /etc/initt*"); will expand \*(C`/etc/initt*\*(C' to \*(C`/etc/inittab\*(C' before running the \*(C`ls\*(C' command. If, on the other hand, you say $status = $proc->start("/bin/ls", "-l", "*"); the \*(C`*\*(C' will stay unexpanded, meaning you'll look for a file with the literal name \*(C`*\*(C' (which is unlikely to exist on your system unless you deliberately create confusingly named files :). For more info on this, look up \*(C`perldoc -f exec\*(C'. If, on the other hand, you want to start a Perl subroutine in the background, simply provide the function reference like $status = $proc->start(\&your_function); or supply an unnamed subroutine: $status = $proc->start( sub { sleep(1) } ); You can also provide additional parameters to be passed to the function: $status = $proc->start(\&printme, "hello", "world"); The start Method returns immediately after starting the specified process in background, i.e. non-blocking mode. It returns 1 if the process has been launched successfully and 0 if not.
The poll method checks if the process is still running $running = $myproc->poll(); and returns 1 if it is, 0 if it's not.
The kill() method: $myproc->kill(); terminates the process by sending it the \s-1SIGTERM\s0 signal. As an option, another signal can be specified. $myproc->kill("SIGUSR1"); sends the \s-1SIGUSR1\s0 signal to the running process. kill returns 1 if it succeeds in sending the signal, 0 if it doesn't.
Set a flag to determine whether the process attached to this object should be killed when the object is destroyed. By default, this flag is set to false. The current value is returned. $current = $proc->kill_on_destroy; $proc->kill_on_destroy(1); # Set flag to true $proc->kill_on_destroy(0); # Set flag to false
Method to set the signal that will be sent to the process when the object is destroyed (Assuming kill_on_destroy is true). Returns the current setting. $current = $proc->signal_on_destroy; $proc->signal_on_destroy("KILL");
Redirects stdout and/or stderr output to a file. Specify undef to leave the stderr/stdout handles of the process alone. # stdout to a file, left stderr unchanged $proc->redirect_output ("/tmp/someapp.stdout", undef);
# stderr to a file, left stdout unchanged $proc->redirect_output (undef, "/tmp/someapp.stderr");
# stdout and stderr to a separate file $proc->redirect_output ("/tmp/someapp.stdout", "/tmp/someapp.stderr"); Call this method before running the start method.
Returns the pid of the forked process associated with this object $pid = $proc->pid;
Returns the start time() of the forked process associated with this object $t0 = $proc->t0();
Returns the stop time() of the forked process associated with this object $t1 = $proc->t1();
Object destructor. This method is called when the object is destroyed (eg with \*(L"undef\*(R" or on exiting perl). If kill_on_destroy is true the process associated with the object is sent the signal_on_destroy signal (\s-1SIGTERM\s0 if undefined).
Returns the exit status of the process as the $! variable indicates. If the process is still running, \*(C`undef\*(C' is returned.
The wait method: $exit_status = $myproc->wait(); waits until the process is done and returns its exit status.
Switches debug messages on and off \*(-- Proc::Simple::debug\|(1) switches them on, Proc::Simple::debug\|(0) keeps Proc::Simple quiet.
Proc::Simple keeps around data of terminated processes, e.g. you can check via \*(C`t0()\*(C' and \*(C`t1()\*(C' how long a process ran, even if it's long gone. Over time, this data keeps occupying more and more memory and if you have a long-running program, you might want to run \*(C`Proc::Simple->cleanup()\*(C' every once in a while to get rid of data pertaining to processes no longer in use.
Please keep in mind that there is no guarantee that the \s-1SIGTERM\s0 signal really terminates a process. Processes can have signal handlers defined that avoid the shutdown. If in doubt, whether a process still exists, check it repeatedly with the poll routine after sending the signal.
If you pass a shell program to Proc::Simple, it'll use \*(C`exec()\*(C' to launch it. As noted in Perl's \*(C`exec()\*(C' manpage, simple commands for the one-argument version of \*(C`exec()\*(C' will be passed to \*(C`execvp()\*(C' directly, while commands containing characters like \*(C`;\*(C' or \*(C`*\*(C' will be passed to a shell to make sure those get the shell expansion treatment.
This has the interesting side effect that if you launch something like
$p->start("./womper *");
then you'll see two processes in your process list:
$ ps auxww | grep womper mschilli 9126 11:21 0:00 sh -c ./womper * mschilli 9127 11:21 0:00 /usr/local/bin/perl -w ./womper ...
A regular \*(C`kill()\*(C' on the process \s-1PID\s0 would only kill the first process, but Proc::Simple's \*(C`kill()\*(C' will use a negative signal and send it to the first process (9126). Since it has marked the process as a process group leader when it created it previously (via setsid()), this will cause both processes above to receive the signal sent by \*(C`kill()\*(C'.
Tim Jenness <[email protected]>
did kill_on_destroy/signal_on_destroy/pid
Mark R. Southern <[email protected]>
worked on \s-1EXIT_STATUS\s0 tracking
Tobias Jahn <[email protected]>
added redirection to stdout/stderr
Clauss Strauch <[email protected]> suggested the multi-arg start()-methods.
Chip Capelik contributed a patch with the wait() method.
Jeff Holt provided a patch for time tracking with t0() and t1().
Brad Cavanagh fixed \s-1RT33440\s0 (unreliable $?)
1996, Mike Schilli <[email protected]>
Copyright 1996-2011 by Mike Schilli, all rights reserved. This program is free software, you can redistribute it and/or modify it under the same terms as Perl itself.