Re: new run-tests

zoe schrieb: > Greg Beaver wrote: >> Hi, >> >> I'm looking at using the new run-tests implementation inside Pyrus, and >> am a bit concerned at some of the assumptions. It seems that all of the >> configuration code assumes that it is running from an application where >> it has absolute control over $argv and friends. A much better design >> would be to separate these dependencies, write some classes whose sole >> purpose is parsing $argv to set up the configuration, and let other >> programs like Pyrus do their own configuration setup and pass that in. >> >> As it is designed now, I have to extend rtWinConfiguration and >> rtUnixConfiguration just to stop it from trying to parse $argv. >> > Hi Greg - I think you are probably right, but I our primary focus was > (is) on writing something that met the requirements > (http://wiki.php.net/qa/runtests/requirements), so I see your > requirement as additional to those. In fact - I don't think it's _that_ > hard to do, but it does need a little thought. I'm happy to add it as an > additional requirement, but I think we have to address it after we have > the basic stuff done. > > Stefan - thoughts? This is definitely an area that needs some work, but for now the main focus should be to "just" write a command line application. The software architecture of run-tests will allow us to add other front ends later. We also still have some (or maybe even a lot) of dependencies on the actual system environment in the code, which we have to get rid of and use objects representing the system environment and the input data. Before we haven't done that, abstracting from $argv (which of course is a good idea and something that we should do on the longer run) doesn't really make much sense imho. Regards, Stefan