ATF work

[Garrett Cooper]

On Fri, Apr 4, 2014 at 12:19 PM, Alan Somers <asomers at freebsd.org> wrote:
> On Fri, Apr 4, 2014 at 1:05 PM, Julio Merino <jmmv at freebsd.org> wrote:
>> Just a couple of minor comments:
>>
>> On Wed, Apr 02, 2014 at 11:07:04AM -0700, Garrett Cooper wrote:
>>> (Just to fill in some context on some of the items here)
>>>
>>> On Wed, Apr 2, 2014 at 9:31 AM, Alan Somers <asomers at freebsd.org> wrote:
>>> > lib/libc/tests/net/Makefile has one tested comment out with the
>>> > comment "test uses rump".  Would it be possible instead to leave the
>>> > test in the build, but put "require.progs rump_server" in the relevant
>>> > test cases' heads?
>>>
>>> Probably carryover from NetBSD that should be pushed back to NetBSD.
>>
>> Why?  rump is "standard" in NetBSD so that's probably not going to fly.
>> It'd be like saying "require.progs = ls".
>
> I don't know about NetBSD, but FreeBSD has a lot of optional stuff in
> base that's compiled in by default.  For example, the entire Bluetooth
> stack can be disabled by WITHOUT_BLUETOOTH=yes make buildworld.  At
> $WORK, I disabled a whole bunch of stuff that way to slim down our
> product's image.  A deeply embedded system, I'm sure, would disable
> even more.  If rump can be disabled in a NetBSD build, then it would
> be worthwhile for rump-based ATF tests to identify themselves via
> require.progs.

    Rump is optional, so I think that Alan's point is valid (from
http://netbsd.gw.com/cgi-bin/man-cgi?mk.conf++NetBSD-current):

     MKRUMP         Can be set to ``yes'' or ``no''.  Indicates whether the
                    rump(3) headers, libraries and programs are to be
                    installed.

                    Default: ``yes''

    I'll add that to my fork.
Thanks!
-Garrett