Makefile.inc1.patch

[Garrett Cooper]

> On Jan 24, 2014, at 8:58, Alan Somers <asomers at freebsd.org> wrote:
> 
>> On Thu, Jan 23, 2014 at 3:21 PM, Alan Somers <asomers at freebsd.org> wrote:
>>> On Thu, Jan 23, 2014 at 3:15 PM, Garrett Cooper <yaneurabeya at gmail.com> wrote:
>>> On Jan 23, 2014, at 2:11 PM, Simon J. Gerraty <sjg at juniper.net> wrote:
>>> 
>>>>>> For options.mk I allow MK_* to already be set and WITHOUT_* to take
>>>>>> precedence over WITH_*.  I also allow makefiles to have their own =
>>>>> lists
>>>>>> of options - separate the policy from the mechanism.
>>>>> 
>>>>> Would that fix this case though?
>>>> 
>>>> I imagine it would make fixing it easier.
>>>> 
>>>>>> I guess you could even allow a per-knob setting as to which takes
>>>>>> precedence.=20
>>>>> 
>>>>> You mean override the default so WITH_* overrides WITHOUT_*?
>>>> 
>>>> Yes - I expect that would be rare, but worth it for completness.
>>>> The important thing is a simple precidence rule.
>>>> 
>>>>>> By simply allowing WITHOUT_* to overrule WITH_*, the Makefile.inc1 =
>>>>> usage
>>>>>> would be greatly simplified.
>>>>> 
>>>>> Maybe=85 the -DNO_* logic is a bit messy=85
>>>> 
>>>> NO_* always wins, it allows a makefile to say "I don't care what you
>>>> want I cannot do that".
>>>> 
>>>> Most places you see -DNO_* used could be -DWITHOUT_* if the semantics
>>>> were not broken as previously described.
>>>> NO_* should be mainly for makefiles to set - like NO_MAN= (i don't got
>>>> no man page man)
>>>> 
>>>>> Curious to see what you have in mind :)..
>>>> 
>>>> Look at contrib/bmake/Makefile
>>> 
>>> Ok, I’ll definitely look at that.
>>> 
>>> Alan,
>>> 
>>> As far as fixing your issue is concerned though, has a fix already been committed or does one still need to be committed? If the latter, does this suffice for today ― with the intent that it will get ripped out in favor of something cleaner in the [near] future?
>> 
>> Umm, I accidentally committed my earlier patch along with a different
>> change today.  Oops.  I'm currently testing your latest patch.  I'm
>> happy with committing it if it works.
>> 
>> -Alan
> 
> Works for me.  Do you want me to commit it for you?

Please!

The only thing I might do is add a comment stating that WITH_ATF is there just to ensure that lib/atf is built in make libraries, so any code which needs to link against it later on when -DNO_TESTS, can.

Thanks!
-Garrett