System waiting for launched process AND forked processes

Hello,

We have written a small client-server program in c/socket. To launch the
server, we call the program which internally fork. The original process
returns immediately but the forked process stays alive and listen to
incoming socket connection.
It’s all fine when running this from bash (we’re on ubuntu) but when
using ruby (with either system, backtick or #x) it’s blocking until both
the original process and the forked process terminate.

Is it what we should expect when calling system? If yes, how could we
work around that?

Thanks

It’s all fine when running this from bash (we’re on ubuntu) but when
using ruby (with either system, backtick or #x) it’s blocking until both
the original process and the forked process terminate.

Is it what we should expect when calling system? If yes, how could we
work around that?

Mine doesn’t seem to do that.

rdp@li49-39:~$ cat spawn.rb
Process.fork {
puts ‘in daemon’
puts Process.pid
sleep
}
puts ‘terminating’

rdp@li49-39:~$ cat spawn2.rb
system("~/installs/ruby_trunk_installed/bin/ruby spawn.rb")
puts ‘done’

rdp@li49-39:~$ ruby -v spawn2.rb
ruby 1.8.6 (2009-3-4 mbari 8B/0x8770 on patchlevel 287) [i686-linux]
terminating
done
in daemon
3199
rdp@li49-39:~$ ps -ef | grep 3199
rdp 3199 1 0 17:41 pts/0 00:00:00
/home/rdp/installs/ruby_trunk_installed/bin/ruby spawn.rb
rdp 3272 477 0 17:41 pts/0 00:00:00 grep 3199

(3199 is still alive).

Perhaps your other process.wait’ing on the first or something?

One thing that might help (in 1.9)
Process.daemon {

something

}

or

Process.fork {
Thread.new { system(“long running command”)}
}

or what not.

GL.
-rp

On 02/17/2010 06:22 PM, Pierre Morel wrote:

We have written a small client-server program in c/socket. To launch the
server, we call the program which internally fork. The original process
returns immediately but the forked process stays alive and listen to
incoming socket connection.
It’s all fine when running this from bash (we’re on ubuntu) but when
using ruby (with either system, backtick or #x) it’s blocking until both
the original process and the forked process terminate.

Is it what we should expect when calling system? If yes, how could we
work around that?

Hmmm, might be that you have an issue because your child’s child is
attached to your ruby process once the starter process stops. There is
a whole range of other possible issues causing this which is hard to
sort out given the little information we have. Did you take proper
measures to demonize your child process after fork?

Kind regards

robert

On 02/17/2010 06:42 PM, Roger P. wrote:

Process.fork {
Thread.new { system(“long running command”)}
}

What do you create the thread for? Also exec is probably a better
choice in this case.

Kind regards

robert

Process.fork {
Thread.new { system(“long running command”)}
}

What do you create the thread for? Also exec is probably a better
choice in this case.

ahh yes exec would work well.
-r

Robert K. wrote:

Hmmm, might be that you have an issue because your child’s child is
attached to your ruby process once the starter process stops. There is
a whole range of other possible issues causing this which is hard to
sort out given the little information we have. Did you take proper
measures to demonize your child process after fork?

Kind regards

robert

I didn’t take any measure to daemonize the child… Now I have done it
(a call to setsid and few other things, I used the example at
http://www-theorie.physik.unizh.ch/~dpotter/howto/daemonize) and it’s
working perfectly.

My mistake was assume that it would work from ruby if it was working
fine from the shell or even from another C++ program.

Thank you very much for your quick response.

Pierre