In the Test::Trap-Code I can't find any workarounds for the test-exits, so I'm as stupid as before :?

Here's what I came up with:

use Test::More 'no_plan';

open my $mail_pipe, '|-', 'cat > /tmp/pm' or die "Can't pipe: $!";
print $mail_pipe 'Test run: ', scalar localtime, "\n";

Test::More->builder->output( $mail_pipe );
Test::More->builder->failure_output( $mail_pipe );

ok( 1, 'one is true' );
ok( ! 0, 'zero is false' );
ok( 'roses', 'roses are read' );
ok( die(), 'violets are blue' );
ok( 'sugar', 'sugar is sweet' );
[download]

The output from this when run is simply "Died at ...". Then /tmp/pm contains this:

Test run: Thu Jun 14 10:19:57 2007
ok 1 - one is true
ok 2 - zero is false
ok 3 - roses are read
1..3
# Looks like your test died just after 3.
[download]

In your case, instead of cat > /tmp/pm, open a pipe to something that will send the mail you want to send (such as 'mail' or 'sendmail'). When the exit happens, that pipe gets closed, and the mail gets sent. The only problem you might have is that if your mail sending program bombs, you'll never know it.

I've to do more than send all Test as a simple mail, but I've solved like that (and I'm not that happy with it):

our $OUTPUT = '';
my $fh = IO::Handle->new();
open($fh, '>', \$OUTPUT);

# Catch all output
Test::More->builder->output($fh);
Test::More->builder->failure_output($fh);
Test::More->builder->todo_output($fh);

..testcode..
BAIL_OUT(..or other exit...)

END {
# This is called even after exit(), so I can access the output
print STDOUT $OUTPUT;
}
[download]