Am Wednesday 15 August 2012 schrieb Daniel Stenberg: > On Tue, 14 Aug 2012, Tim Ruehsen wrote: > >> It shares no code with current Wget, AFAICT. > > > > 90% correct. I already rewrote the basic parts for Mget, so a big bunch > > of work is done. > > I'm far from sure about that. You rewrote significant portions of a 15+ > years old project with lots of "proven in use" legacy and with a serious > lack of actual good test cases.
15+ years is the best argument for a rewrite. Wget's basic functionality is quite simple. But you're right: it took 15+ year to add some add-ons and tweak for all the little corner cases. And I am absolutely on your's and Micah's side regarding the test cases. > I strongly agree with Micah's previous vision about getting a serious test > suite setup *first* to make sure and prove how wget1 is working right now > and then use that to make sure that any considered work on wget2 runs the > same (or better) with the use of this test suite. > > IMHO, projects without (decent) test suites cannot do large rewrites > without getting seriously injured by regressions. This is of course just > my own opinion. It would be perfect, to have a large test suite. If someone works out a test suite design for wget1, I would spend some time into the coding. Once finished, the test suite should be good for the wget2 thing... Any volunteers to create a draft for a test suite design ? I know there is a test suite, but how usable is it, what is missing, what is needed ? Could anyone give us an overview ? Tim
