package Test::Fork;
use strict;
use warnings;
our $VERSION = '0.02';
use base 'Test::Builder::Module';
our @EXPORT = qw(fork_ok);
my $CLASS = __PACKAGE__;
sub note {
my $msg = shift;
my $fh = $CLASS->builder->output;
print $fh "# $msg\n";
}
=head1 NAME
Test::Fork - test code which forks
=head1 SYNOPSIS
use Test::More tests => 4;
use Test::Fork;
fork_ok(2, sub{
pass("Test in the child process");
pass("Another test in the child process");
});
pass("Test in the parent");
=head1 DESCRIPTION
B<THIS IS ALPHA CODE!> The implementation is unreliable and the interface
is subject to change.
Because each test has a number associated with it, testing code which forks
is problematic. Coordinating the test number amongst the parent and child
processes is complicated. Test::Fork provides a function to smooth over
the complications.
=head2 Functions
Each function is exported by default.
=head3 B<fork_ok>
my $child_pid = fork_ok( $num_tests, sub {
...child test code...
});
Runs the given child test code in a forked process. Returns the pid of the
forked child process, or false if the fork fails.
$num_tests is the number of tests in your child test code.
Consider it to be a sub-plan.
fork_ok() itself is a test, if the fork fails it will fail. fork_ok()
test does not count towards your $num_tests.
# This is three tests.
fork_ok( 2, sub {
is $foo, $bar;
ok Something->method;
});
The children are automatically reaped.
=cut
my %Reaped;
my %Running_Children;
my $Is_Child = 0;
sub fork_ok ($&) {
my($num_tests, $child_sub) = @_;
my $tb = $CLASS->builder;
my $pid = fork;
# Failed fork
if( !defined $pid ) {
return $tb->ok(0, "fork() failed: $!");
}
# Parent
elsif( $pid ) {
# Avoid race condition where child has run and is reaped before
# parent even runs.
$Running_Children{$pid} = 1 unless $Reaped{$pid};
$tb->use_numbers(0);
$tb->current_test($tb->current_test + $num_tests);
$tb->ok(1, "fork() succeeded, child pid $pid");
return $pid;
}
# Child
$Is_Child = 1;
$tb->use_numbers(0);
$tb->no_ending(1);
note("Running child pid $$");
$child_sub->();
exit;
}
END {
while( !$Is_Child and keys %Running_Children ) {
note("reaper($$) waiting on @{[keys %Running_Children]}");
_check_kids();
_reaper();
}
}
sub _check_kids {
for my $child (keys %Running_Children) {
delete $Running_Children{$child} if $Reaped{$child};
delete $Running_Children{$child} unless kill 0, $child;
note("Child $child already reaped");
}
}
sub _reaper {
local $?; # wait sets $?
my $child_pid = wait;
$Reaped{$child_pid}++;
delete $Running_Children{$child_pid};
note("child $child_pid reaped");
$CLASS->builder->use_numbers(1) unless keys %Running_Children;
return $child_pid == -1 ? 0 : 1;
}
$SIG{CHLD} = \&_reaper;
=head1 CAVEATS
The failure of tests in a child process cannot be detected by the parent.
Therefore, the normal end-of-test reporting done by Test::Builder will
not notice failed child tests.
Test::Fork turns off test numbering in order to avoid test counter
coordination issues. It turns it back on once the children are done
running.
Test::Fork will wait for all your child processes to complete at the end of
the parent process.
=head1 SEE ALSO
L<Test::MultiFork>
=head1 AUTHOR
Michael G Schwern E<lt>schwern@pobox.comE<gt>
=head1 BUGS and FEEDBACK
Please send all bugs and feature requests to
I<bug-Test-Fork> at I<rt.cpan.org> or use the web interface via
L<http://rt.cpan.org>.
If you use it, please send feedback. I like getting feedback.
=head1 COPYRIGHT and LICENSE
Copyright 2007-2008 by Michael G Schwern E<lt>schwern@pobox.comE<gt>.
This program is free software; you can redistribute it and/or
modify it under the same terms as Perl itself.
See F<http://www.perl.com/perl/misc/Artistic.html>
=cut
42;