[Mesa-dev] Current tinderbox regression (egl, parallel make)

[Chia-I Wu]

On Tue, Apr 27, 2010 at 6:14 AM, Dan Nicholson <dbn.lists at gmail.com> wrote:
> On Mon, Apr 26, 2010 at 2:16 PM, Chris Ball <cjb at laptop.org> wrote:
>> http://tinderbox.x.org/builds/2010-04-26-0020/logs/libGL/#build
>> gmake[3]: *** No rule to make target `depend', needed by `default'.  Stop.
>> gmake[2]: *** [../../../progs/egl/eglut/libeglut-screen.a] Error 2
>> gmake[2]: *** Waiting for unfinished jobs....
>> [...]
>> ar: creating libeglut-x11.a
>> ar: creating libeglut-screen.a
>> make[1]: *** [subdirs] Error 1
>> make: *** [default] Error 1
>> Seems to have broken in the last twelve hours or so, building with -j4.
>> It's repeatable across build attempts.
> progs/egl/opengl/Makefile is really nasty and makes this tough to
> debug. I can't tell exactly what causes the error, but the build is
> trying to jump from egl/opengl to egl/eglut to build the two
> convenience libraries. Why it sees the depend target but not the rule,
> I don't know. Does it build without parallel jobs?
The error is not trivial to me either.  Probably because two instances
of make jump to egl/eglut for different dependencies at the same time,
and/or the use of $(wildcard *.c) in eglut.

Anyway, I've pushed a fix that add egl/eglut to PROGRAM_DIRS and build
it explicitly before the other EGL demos.  Hopefully, it fixes the
issue.
> It would be much better if egl/eglut was just built first before going
> to build the demos. Adding Chia-I since I can't really tell what's
> supposed to be happening.

-- 
olv at LunarG.com