Al Tobey wrote:
Sysadmins everywhere feel this "broken tests are a good thing"
syndrome as real, almost physical, pain nearly every time they work
with CPAN these days. It's great that TDD is making the progress it
has, but I think some coders got religion and missed the point:
quality.
Maybe the catch phrase is "every time a CPAN test fails, a sysadmin
loses a day off the end of his life." It really is that frustrating
at times.
Great point, Al.
One of the surprising realizations (to me at least) from the Vanilla
Perl project is how many modules fail tests because of portability
problems in the tests, rather than the code. See this page for a list
of some of the issues highlighted so far:
http://win32.perl.org/wiki/index.php?title=Vanilla_Perl_Problem_Modules
Many of the test failures can be attributed to:
* non-portable path expectations
* forks and process control
* assumptions of unix command tool availability
* signal handling
I think authors need to aim to have the quality of test code be the same
as the quality of module code. (Though I'll admit that I don't always
live up to that standard myself.)
Regards,
David Golden
