Devel::Cover has been giving Test::More some indigestion lately. Threads. Bleh. Paul put out a new version so I figured I'd try it out. It faired reasonably well, everything but the threaded stuff ran better than the last version.
So I looked at the coverage output. ---------------------------- ------ ------ ------ ------ ------ ------ ------ File stmt branch cond sub pod time total ---------------------------- ------ ------ ------ ------ ------ ------ ------ blib/lib/Test/Builder.pm 92.3 79.7 66.7 96.6 100.0 76.5 87.0 blib/lib/Test/More.pm 95.0 90.6 82.2 94.4 95.2 22.8 92.9 blib/lib/Test/Simple.pm 100.0 n/a n/a 100.0 100.0 0.7 100.0 Total 93.5 83.5 72.8 96.0 98.2 100.0 89.4 Not bad. Dug into Test::More's details. Found a bunch of false negatives. Devel::Cover doesn't seem to like "eval <<HEREDOC". Its not useful to check the 0 || 0 case of "my $class = ref $proto || $proto." Some assertions that should never happen (the WHOA!s). I came across one half of one conditional that wasn't getting tested. So I figured, for completeness sake, I'd test it. Whatdya know, that feature was broken! [1] And undocumented! So even when you approach 100% there's still bugs to be found with simple coverage analysis. [1] For the record, its that the $how_many parameter of skip() and todo_skip() is supposed to be optional when your plan is 'no_plan'. -- Michael G Schwern [EMAIL PROTECTED] http://www.pobox.com/~schwern/ Don't step on my funk
