On Thu, Jul 14, 2005 at 08:37:06PM +0200, Joerg Schilling wrote: > If you install SchilliX on a previously empty box, you get a useable > system. If you allow me to remove all files that are not part of > OpenSolaris, it will turn your box into a door stopper.
Are you claiming that it needs to be possible to test OpenSolaris in isolation? Test design requires you to understand what you're testing; if you've made no changes to something and it's not used in the course of running your tests, there's no reason to be concerned with it. To be more concrete, if you're worried that replacing libm might affect the outcome of the tests you are running against your changes to ls, you need to evaluate whether ls uses libm and whether your changes might have affected the way it is used. If it's not used, or there's no conceivable way your changes to ls might have affected the way ls uses libm's interfaces, there's no reason to include libm as a variable in your test plan for ls. Presumably, you have verified already by use of a focused test suite that SchilliX's libm is 100% compatible with Solaris's anyway, so which is installed on the test system would not normally be of interest when testing ls. This seems like common sense; the alternative would be enormous test matrices for even the most trivial changes. -- Keith M Wesolowski "Sir, we're surrounded!" Solaris Kernel Team "Excellent; we can attack in any direction!" _______________________________________________ opensolaris-discuss mailing list [email protected]
