Makefile.inc1.patch

[Alan Somers]

On Thu, Jan 23, 2014 at 2:34 PM, Garrett Cooper <yaneurabeya at gmail.com> wrote:
> Ugh. Backwards logic (sorry)...
>
> On Jan 23, 2014, at 1:32 PM, Garrett Cooper <yaneurabeya at gmail.com> wrote:
>
>> On Jan 23, 2014, at 1:29 PM, Alan Somers <asomers at freebsd.org> wrote:
>>
>>> On Thu, Jan 23, 2014 at 2:23 PM, Garrett Cooper <yaneurabeya at gmail.com> wrote:
>>>> On Jan 23, 2014, at 1:11 PM, Alan Somers <asomers at freebsd.org> wrote:
>>>>
>>>>> In that case, I'm missing something.  I can't find any makefiles that
>>>>> reference MK_ATF or a related variable.  What is the effect of setting
>>>>> WITH_ATF ?
>>>>>
>>>>> -Alan
>>>>
>>>>       NO_TESTS forces WITHOUT_TESTS to be set. So, if I set NO_TESTS in the various build steps it will force ATF to not be built. For that reason (and that reason alone) I reintroduced WITH_ATF just for Makefile.inc1 (but you could replace it with something else like WITH_ATF_LIBS, etc, if the naming is too confusing).
>>>> Thanks!
>>>> -Garrett
>>>
>>> I get that much, but what I don't understand is what direct affect
>>> WITH_ATF has.  Did you forget to reintroduce a ".if defined(MK_ATF)"
>>> in some other file?
>>
>> No, the purpose of WITH_ATF is to override NO_TESTS, so building the ATF libs now has two conditions:
>>
>> build_atf_libs = (is WITH_ATF defined?) && (is WITHOUT_TESTS defined?)
>
> build_atf_libs = (is WITH_ATF defined?) && (is WITH_TESTS defined?)

Should that be || instead of && ?

>
>> versus one:
>>
>> build_atf_libs = (is WITHOUT_TESTS defined?)
>
> build_atf_libs = (is WITH_TESTS defined?)
>
>> This allows us pepper NO_TESTS around and thus not build tests in the build process unless they’re _really_ needed (e.g. in make everything).

Ok, I think I get it now.  It's a recursive thing.  At the top level,
your patch adds WITH_ATF to MAKE.  Then, in a child make process, the
presence of WITH_ATF causes _lib_atf to be defined.  Is that correct?

-Alan