Re: We Want Quality

On Thu, 29 Jun 2000, Andi Gutmans wrote: > At 02:09 PM 6/29/00 -0500, Andrei Zmievski wrote: > >On Thu, 29 Jun 2000, Andi Gutmans wrote: > > > From my experience regression tests are crap. They barely catch anything. > > > Having testers run RC's on their development sites (real life scripts) is > > > probably the best regression test you can do. > > > For example, Zend.com's code covers most features PHP has to offer, and > > > when testing a version on that (of course an internal development version > > > of Zend.com) you are much more likely to catch bugs than with a regression > > > test. > > > >True, but regression test can be useful (like for that min()/max() > >stuff). > > Yeah but most developers don't like nor have time to write regression tests > and using that time to fix bugs is more useful. As mentioned before, regression tests are there to prevent a bug from reoccuring. You cannot know in advance what bugs you will create. I'm running automatic builds on a couple of platforms with various targets (Apache, APXS, CGI and if possible, CGI-ZTS). > Seriously, having x people run an RC on their huge web sites is probably > the best test. We just need to make sure we cover sites that include MySQL, > Postgres, ODBC, Oracle, Linux, Slowlaris, FreeBSD and so on. > I think that would work and it would be much much easier to find > volunteers. And if a volunteer doesn't have time, we send a hit man :) Well, just add redundancy. Let multiple people take care of one target, for example. - Sascha