Question

In my Perl/CGI web application, I sometimes need to run a long process which makes the wait for the next page interminable. So I’ve been disabling the buffer as below so that the page contents get sent before the long process runs.

local $| = 1;
print "Content-type: text/html\n\n";
print $output;
&background_process();

However it seems to me that the buffer has its uses and I should not be in the habit of doing this. Is there a better way to run a long process and still return html to the client quickly? Should I be forking or somesuch?

Answer 1

Here is some code I use to spawn off a background process in my FastCGI script:

$SIG{CHLD} = 'IGNORE';

# This should flush stdout.
my $ofh = select(STDOUT);$| = 1;select $ofh;

my $kpid = fork;
if ($kpid)
{
    # Parent process
    waitpid($kpid, 0);
}
else
{
    close STDIN;
    close STDOUT;
    close STDERR;
    setsid();
    my $gpid = fork;
    if (!$gpid)
    {
        open(STDIN, "</dev/null") ;
        open(STDOUT, ">/dev/null") ;
        open(STDERR, ">/dev/null") ;
        # Child process
        exec($pgm, @execargs) ;
    }
    exit 0;
}

A couple of salient points here:

• I ignore SIGCHLD, which is the signal
  you get when your child process
  exits.
• I flush STDOUT, so the web browser gets the whole page.
• I fork twice to make sure that the web server will be able to exit even if the process I’m trying to spawn is still running.
• I redirect all the standard file descriptors to make sure that the grandchild process is fully detached from the parent.