Re: better tests for feature test macro permutations?

[Roland McGrath <roland at hack dot frob dot com>]

I agree about the utility of running tests in a situation closer to a real
installed system.  However, I think it's also worthwhile to keep the
current method, which doesn't require a big install step or other such
stuff to run a new test you've just added or repeat tests for things you've
just changed.

Back to the header/macro test subject, I do think that a fairly exhaustive
test arrangement made mostly automagic is the right way to go.  Each subdir
Makefile sets "headers" and I don't think it would be very hard to use that
to drive automagic per-header tests in each subdir (i.e. implicitly in
Rules).  For the cases that are supposed to #error, it's pretty
straightforward to use variables of formulaic computed names, e.g.:

feature-error-foo.h := _POSIX_SOURCE __STRICT_ANSI__

Off hand the machinery to do all this doesn't seem like it should be
particularly hard.  Not that I intend to put any effort into it myself any
time soon.  But if someone wants to take this on, I'd be glad to give
advice about the makefile magic.


Thanks,
Roland