SA Bugzilla – Bug 1489
Need more regression tests
Last modified: 2004-02-28 13:58:20 UTC
While working on cleaning up rule-related files for a 2.50 release, I see that the current regression-tests only test 34 rules of our total 844. To me, this isn't very useful. So for 2.60, we should either come up with more regression tests or just dump the idea of rule testing in this way.
Subject: Re: [SAdev] New: Need more regression tests > So for 2.60, we should either come up with more regression tests or just dump > the idea of rule testing in this way. argh. we really should write more, but it's painful. ;) I strongly dislike the idea of dumping them BTW, I've found they've shown up bugs on several occasions...
Regression tests are most desireable for complicated rules. Most of the tests do indeed test complicated stuff and it's good to have those regression tests. I also don't think the lack of testing is a valid reason for crapping the tests that we do have.
Subject: Re: Need more regression tests On Sun, Feb 16, 2003 at 04:19:22PM -0800, bugzilla-daemon@hughes-family.org wrote: > I strongly dislike the idea of dumping them BTW, I've found they've shown > up bugs on several occasions... That's fine. I don't think it hurts anything to have them, but only testing 22 (after removing obsolete rules) of 844 rules is pretty pathetic. I figure if you're not going to go "all the way" you may as well not bother. ;) We'll have to work on adding more for 2.60 then. :)
Subject: Re: [SAdev] Need more regression tests > That's fine. I don't think it hurts anything to have them, but only > testing 22 (after removing obsolete rules) of 844 rules is pretty > pathetic. I figure if you're not going to go "all the way" you may as > well not bother. ;) Well... I guess processing time is all it hurts; rough estimate: 1 make test a day per developer * 10 developers * 365 days / year * 3 seconds + 500 CPAN installs * 3 seconds = 3.5 hours a year Okay... so maybe it's not so important. Regardless, we don't need tests for rules like: body VIAGRA /VIAGRA/i > We'll have to work on adding more for 2.60 then. :) Only if necessary, please :-)
hmm, we didn't make much progress on this. bad developers ;) we need more testing in general IMO...
closing bug. again, this is a statement of opinion -- albeit one that we all share. Let's just add it to a a coding style guide or something instead of wasting bugzilla space ;)
Subject: Re: Need more regression tests On Sat, Feb 28, 2004 at 02:55:01PM -0800, bugzilla-daemon@bugzilla.spamassassin.org wrote: > closing bug. again, this is a statement of opinion -- albeit one that we all > share. Let's just add it to a a coding style guide or something instead of > wasting bugzilla space ;) Well, less statement of opinion than "ongoing issue". And it's still "wasting" bugzilla space, just not visible on the web page. ;)
'Well, less statement of opinion than "ongoing issue". And it's still "wasting" bugzilla space, just not visible on the web page. ;)' yeah whatever ;) wasting my attention span, then!
Theo, did you see this? http://www.pathname.com/~quinlan/test_cover_db/coverage.html
Subject: Re: Need more regression tests On Sat, Feb 28, 2004 at 10:58:21PM -0800, bugzilla-daemon@bugzilla.spamassassin.org wrote: > Theo, did you see this? > > http://www.pathname.com/~quinlan/test_cover_db/coverage.html Yeah, I did. really neat! :) I tried using Devel::Cover, but it failed HORRIBLY in "make test" (>95% failure rate). After poking around a bit, it seems like it requires perl 5.8 (even though it says 5.6.1 is fine). So I gave up trying. :(
Subject: Re: Need more regression tests I'm using 5.6.1. I only had a few warnings about it not being able to find some modules (especially in bayesdbm.t IIRC). It definitely points out some areas where we need better coverage. (Hence, I wrote the new DNS tests... which fail due to some unknown issue with the Received header and/or trusted Received header code.