locking solved

[raptor]

hi,

here is my final solution..... of my locking-oddisey :")
see the scripts below....
oops .. the update is triggered only if the commit message beings with -!-
(exclamation mark)


thanx for the help
=====
iVAN
address@hidden
=====


#!/usr/bin/perl
#Author : raptor
use strict;
my $lockfile = '/var/lock/cvs/update.lck';
my $userName = getpwuid($<);
my $updater = '/path/to/updater.pl';

 #get the directory we are working on
my ($dir) = $ARGV[0] =~ /^([^\s]+?)\s/;

sub isLocked { -e $lockfile };
sub lock { open FILE, ">$lockfile" or die "Can't create lock file\n" };

#================= MAIN
 #get the all info
undef $/;
my $msg = <STDIN>;
 #now check to see if the log message begins with '!'
unless ( $msg =~ /Log Message:.!/gs) {
    print qq{$userName : $dir
$msg
--------
    };
}# we will do our actions then
else {
 print "loginfo : -!- used will do update ASAP.....\n";
 unless (isLocked()) { print "loginfo: scheduling update...\n"; lock() }
    else { print "loginfo : update already scheduled, it is ok don't
worry...\n"};
};



__END__

=head1 DESCRIPTION

 Use those scripts so that U can have your main Development
 working copy up to date. Primary usefull for Web development.
 The first script is attached as log-info script.(see: 'info cvs')
 What it does is to create a file whenever commit happen, if
 this file is already created it does nothing.
 The second script stays up all the time and on every X seconds
 checks for this file...
 if there is no file it just sleeps X seconds, otherwise it checks
 if there are some 'cvs'-processes hanging around and if yes
 it sleeps again.
 But if there is no such processes it makes an cvs-update.

 Why so much trouble ?
 'Cause 'cvs' work on a per directory basis i.e. when u make commit
 it fires log-info script for every directory that has to be updated
 (i.e cvs unlocks only the current cvs-dir on which your log-info
  script has been started), so if u want to make cvs-update inside
  the log-info script u can't 'cause other directories are locked.
  I tried to start cvs-update process via both exec & fork, but the
  cvs processes was still hanging around waiting for me to finish.
  (Can figure out other way how to run this as stand-alone process!!)
  The main problem is to broke parent-child reltionship.

 One other solution would be to update only the directory on which
 log-info script has been called, which i'm figuring just now
 (as I write this docs). This probably will be better solution !!
 This will fire too much standalone/outside cvs-updates,
 don't know which is worst. One-big or many-small!!

=head2 TODO

 - make the second variant i just explained
 - make the updater script like a daemon
 - create init.d script so it can be executed on bootup

=cut
===================================updater.pl==========================
#!/usr/bin/perl
#Author : raptor
use strict;
my $userName = getpwuid($<);
my $cvs = '/usr/bin/cvs';
my $repo = '/path/to/repository';
my $lockfile = '/var/lock/cvs/update.lck';
my $workcopy = '/cvstest/mytest';

sub isLocked { -e $lockfile };
sub cvsActive { my $pids = `pidof cvs`; chop $pids; $pids };
sub unlock { unlink $lockfile };

sub update {
    chdir $workcopy;
    system "$cvs -d $repo update -d -A";
    print 'ERROR: Can not update the working copy' if $? > 0;
};

#================= MAIN
print "Process-ID : $$\n";
while ( 1 ) {
 print '.';
 if (isLocked() && ! cvsActive() ) {
    update(); unlock();
 };
 `sleep 5`;#using perl sleep didn't worked !?!!!
};


__END__