AMDG vicente.botet wrote:
It's definitely impractical to test every possible pair. Even if it is restricted to stay within a single library, this would be many thousands of compilations...
Have someone measure the time that it will take already? Can this kind of test runed once a week?
The units library alone contains 291 public headers. That's 42,195 pairs to compile. For comparison, the regression tests generate 23,160 targets on como--including making directories.
I don't know very well bjam, but can the following variation of you Jamfile compile in include_all_rev the same files than in include_all but included in reverse order. If yes at least this test could improve the test coverage.
I don't know how much that will actually help, given that the headers include each other. I'll update the Jamfile though, as it can't hurt... In Christ, Steven Watanabe