If you were using Test::More, this would be easy and documented. The output handle is set by the internal Test::Builder object which itself has the methods ->output( $fh ), ->failure_output( $fh ), and ->todo_output( $fh ) to set the various filehandles.

Here is an example from one of my scripts which shows test output being captured.

use Test::More;
use vars qw( $TEST_OUTPUT );

main( @ARGV );
exit 0;

sub main
{
    intialize( @_ );

    ok( ... );

    print $TEST_OUTPUT;

    1;
}

sub initialize
{
    $| = 1;

    open my $fh, ">", \ $TEST_OUTPUT or die;
    my $b = Test::More->builder;
    $b->output( $fh );
    $b->failure_output( $fh );
    $b->todo_output( $fh );
    1;
}
[download]