perl pgrep returning multiple pids - perl

Question

I have some a perl code like this to get the pid of the process if it
existed. I wanted to know the return status of pgrep so I can find out if the pid is valid or not. Here is the code:
#!/usr/bin/perl -w
my $pid=`pgrep -n -f \"akonadiserver\"`;
print "$? ";
chomp $pid;
print "Output1: $pid\n";
$pid=`pgrep -n -f \"akonadiserver\"`;
my $evalue=${^CHILD_ERROR_NATIVE};
chomp $pid;
print "Output2: $pid, $evalue\n";
But whenever I run it, I get the following:
0 Output1: 3054
Output2: 3056, 0
another run
0 Output1: 3059
Output2: 3061, 0
But in the system:
$ pgrep -n -f akonadiserver
1862
Several problems seen here:
1) The pid returned is changing all the time, most likely it is matching
the grep process itself.
2) I don't have a way to find out of the pid returned is valid or not. $?
doesn't help.
How do I check the return status and how do I use pgrep properly in perl?

Answer 1

Running `pgrep -n -f akonadiserver` from a Perl script, is equivalent to running sh -c "pgrep -n -f akonadiserver" from the terminal.
From perlop :
qx/STRING/:
A string which is (possibly) interpolated and then executed as a system command with /bin/sh or its equivalent.
The problem is that using the -f option with pgrep will match against the full command line used to invoke the process. This will also include the current pgrep command, since its command line will look like sh -c "pgrep -n -f akonadiserver" which will match the string "akonadiserver". Since you also use the -n flag with pgrep it will only return the newest process that matches "akonadiserver" and this will most likely be the pgrep command itself (sh -c "pgrep -n -f akonadiserver").. So that explains why you get the wrong process id.
It does not however explain why it still works from the terminal window. The reason is that from the terminal window, the pgrep command is run with bash and not sh (as from the Perl script), see this question for more information about the difference: Difference between pgrep in sh and bash. As a result, running pgrep -f from the terminal window will not include the pgrep command itself, and therefore it will work for your case.
To obtain a reliable result from Perl, you should use Proc::ProcessTable:
use feature qw(say);
use strict;
use warnings;
use Proc::ProcessTable;
my $pat = "akonadiserver";
my $pt = Proc::ProcessTable->new();
my #tab;
for my $p ( #{$pt->table} ){
if ($p->cmndline =~ /\Q$pat\E/) {
push #tab, {pid => $p->pid, start => $p->start};
}
}
say $_->{pid} for (sort {$b->{start} <=> $a->{start}} #tab);

Answer 2

It seems to me that pgrep returns the pid of the newest (-n) child process started by the process you are looking for, at least that is what happens on my system.
I have always used ps for such things, how about:
my $pid = (split(" ",`ps -A | grep \"akonadiserver\"`))[0];
instead, or
my #pids = map{ $_ =~ /(\d+)/; $1 }split("\n", `ps -A | grep \"akonadiserver\"` );
print "mutiple instances: #pids\n" if #pids > 1;
print "pid: ".$pids[0]."\n" if #pids == 1;
to get all instances running into an array and then do what you want to with it.