Quoting Bryce Harrington <bryce@...260...>:
If someone could narrow that to a list of no more than, say, 20 cases, it would be feasible to automate running as a weekly check.
20 test cases would mean roughly ... log2(20) ~= 4.3 ... distinct macros.
(I'm assuming you didn't mean 20 individual macros to test, which would mean testing ... 2^20 = 1048576 ... different permutations)
This is, of course, assuming that the value of the #define isn't itself significant...
A lot of these things are tied to platform feature tests, so we kind of have to ignore those, and some are repeated tests for the same macro, but ... even given that I don't easily forsee being able to pare ~2000 some conditions down to four or five to actually test.
Oh. You know what, I bet a lot of those are header file #include armor. That's the one GOOD use of #ifndef.
Can you subtract the number of header files in the tree from both your #ifndef and #define figures and see what you get?
-mental