Question

I have application that works using Perl’s CGI::Fast.

Basically mainloop of the code is:

while (my $cgi = CGI::Fast->new() ) {   do_whatever( $cgi ); } 

Now, I wanted to add ability to kill it, but without disrupting currently processed request. To do so, I added handling to SIGHUP. More or less along the lines:

my $sighupped = 0; $SIG{ 'HUP' } = sub { $sighupped = 1; }; while (my $cgi = CGI::Fast->new() ) {   do_whatever( $cgi );   exit if $sighupped; } 

Which works really great when it comes to ‘not disturbing process while it handles user request’. But, if the process is waiting for new request the sighup-based exit will not be executed until it will finally get some request and process it.

It there any workaround for this? What I would like to achieve it to make the script exit immediately if the HUP (or other signal, it can be changed) reaches it while waiting for request.

Answer 1

You could use the block version of the eval function and the alarm function to add a timeout to Fast::CGI. In the code below I have set a five second timeout. If it times out it will go back to the start of the loop which will check to see if $continue has been set to zero yet. If it hasn’t then it we start a new CGI::Fast object. This means that a maximum of five seconds will go by after you send a HUP before the code will start to stop if it was waiting for a new CGI::Fast object (the timeout has no affect on the rest of the loop).

my $continue = 1; $SIG{HUP} = sub { $continue = 0 }; while ($continue) {     my $cgi;     eval {         local $SIG{ALRM} = sub { die 'timeout\n' };         alarm 5; #set an alarm for five seconds         $cgi = CGI::Fast->new;         alarm 0; #turn alarm off     };     if ($@) {         next if $@ eq 'timeout\n'; #we will see the HUP signal's change now         #died for a reason other than a timeout, so raise the error         die $@;     }     last unless defined $cgi; #CGI::Fast has asked us to exit     do_stuff($cgi); }  #clean up