On Wed, 27 Oct 2004 23:13:13 +0100, you wrote:
- writing some regression tests to show your code is correct
I still don't really know what I might test here. Maybe try examples of cases where the current code does not act like Windows; there are also some cases where neither acts like windows (but they seem to have the same problems). The problem is I probably wouldn't think of cases where my code causes regressions (otherwise it wouldn't do so:-)), and this isn't really something in which you can enumerate all the different combinations of options (at least without infinite time and patience). Thanks for your comments,
Actually there are already some printf tests in dll/msvcrt/tests/scanf.c (despite its name). If you look at the code you see that some printf tests ( ok( ...) ) are surrounded by a todo_wine { ... }. That means the test is currently failing, we know that and will be noticed when the test "suddenly" succeeds.
Just type "make test" in the dll/msvcrt/tests dir and see if your patch has made any changes. Add more or change if you like. The documentation in the developer guide about tests is good.
IMO this patch should not be delayed committing just because not all possible tests are there yet. Others can help here as well, starting with at least three printf bugs in wine's bugzilla.
Rein.