On Fri May 11 21:05:28 2012, ETHER wrote:
Show quoted text> consider this test:
>
> use strict;
> use warnings;
>
> use Test::More;
> use Test::NoWarnings;
>
> print STDERR "stderr";
>
> done_testing;
>
> It prints:
>
> perl warningstest.t
> stderr1..0
> ok 1 - no warnings
> # Looks like you planned 0 tests but ran 1.
>
> Without a LF, the message to STDERR does not cause a test failure!
And
Show quoted text> yet, if one adds a "\n" to the print, the test *does* fail. At least,
> this is inconsistent; or worse, it is letting warnings leak through
that
Show quoted text> should be caught.
>
There are 3 things going on here.
1 Printing to STDERR is quite different to a warning and you should not
expect Test::NoWarnings to catch this. A warning is a call to warn()
either in the script or from the perl compiler/run time.
2 The test harness expects the test results to be output to STDERR and
expects them to be on their own line. Printing to STDERR without LFs
during a test is very likely to interfere with that and the test harness
may be unable to correctly parse the test results.
3 The test fails with or without the \n for me. It fails even without
the print. Its exit value is 255. I guess this is because
Test::NoWarnings executes a test in an END block, which comes after
done_testing(). This is an error, you cannot use done_testing with
Test::NoWarnings like this.
If you really must use done_testing, you'll have to make sure it runs in
an END block after Test::NoWarnings's END block but be careful. You'll
need to do something messy like
my $done_testing = 0;
test
test
...
test
$done_testing = 1;
END {
done_testing() if $done_testing;
}
but you might have to position the END block carefully to have it run
after Test::NoWarnings END block.
