On Fri, Jul 09, 2004 at 12:10:52PM -0500, Pete Krawczyk wrote:
> Consider the following code:
>
> $impclass ||= implementor($scheme) ||
> do {
> require URI::_foreign;
> $impclass = 'URI::_foreign';
> };
>
> That's in URI.pm, lines 54-58.
>
> Devel::Cover treats that as a conditional. So short of deleting
> URI::_foreign, that do BLOCK is never going to return false.
There's a whole set of these sort of problems.
sub new {
my $proto = shift;
my $class = ref $proto || $proto;
Devel::Cover wants you to consider the case where both ref $proto and $proto
are false. This is an (almost) impossible condition and certainly not an
interesting one to bother testing.
Then there are conditions that can't both occur in a single environment.
if( $^O eq 'MSWin32' ) {
do this;
}
else {
do that;
}
Unless you start combining the coverage runs from all hopolites. Which
might not be a bad thing. It certainly would help things like Class::DBI
which looks like it has piss poor coverage because it has so many optional
modules for testing.
Don't be mesmerized by 100% coverage. Devel::Cover isn't perfect and even
if it was there are uncoverable things. Even if you were to achieve 100%
coverage that doesn't guarantee anything. Coverage doesn't say anything
about the data being tested against, things going funny because the
environment changed, or an external utility (database).
Test coverage is a useful *heuristic* for test effectiveness. Like all
heuristics if you push it too far it falls apart. Get as close to 100% as
is useful and don't worry about the rest.
--
Michael G Schwern [EMAIL PROTECTED] http://www.pobox.com/~schwern/
11. Every old idea will be proposed again with a different name and
a different presentation, regardless of whether it works.
-- RFC 1925
