On Tue, 8 Jan 2002, Dimitrie O. Paun wrote: [...]
Now, from the POV of Wine, we don't care (just like the kernel), if the executable is a native binary or a #! executable. Everything else are a few Makefile rules, which are conditional on the given compiler/interpreter being available (which can be checked quite easily by configure).
That's true on Unix because sh, perl, and C executables will just work. But if some of your tests are sh scripts you will have trouble running them on Windows. We probably won't often need to run all the tests in Windows, but I can imagine that it would still be necessary to check behavior on different setups: in 16bpp vs. 32bpp, in the english vs. the russian vs. chinese version, with IE 5 installed or not installed, etc. So we need a framework that makes it easy to run all the tests on Windows. Since sh scripts tend to invoke a ton of Unix tools like expr, awk, sed, perl, this seems not to be a good basis for writing tests.
But I agree with the approach: a test is an executable that returns 0 if successfull and non-zero if not. It's pretty much the foundation of my proposal except that all tests should either be of the same type: perl or C (or whatever).
-- Francois Gouget fgouget@free.fr http://fgouget.free.fr/ Cahn's Axiom: When all else fails, read the instructions.