Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 04 Apr 2015 18:49:46 -0600
From:      Ian Lepore <ian@freebsd.org>
To:        Garrett Cooper <yaneurabeya@gmail.com>
Cc:        Adrian Chadd <adrian@freebsd.org>, "freebsd-arch@freebsd.org" <freebsd-arch@freebsd.org>
Subject:   Re: make universe failures?
Message-ID:  <1428194986.82583.149.camel@freebsd.org>
In-Reply-To: <BF91F228-A5F7-4553-B578-5050A9D73E16@gmail.com>
References:  <CAJ-VmokOx=5p6veRL%2BiCpBdTCsyDYeNE4Drz-16hUG_b7akwYQ@mail.gmail.com> <AA7A173C-3931-4EB0-BE1B-3C58ACCE10F0@gmail.com> <CAJ-Vmo=aQvGvjBPnd8AvZY5dDD8pq6PKnkCg47NQ58Y_5PS1kQ@mail.gmail.com> <310573CF-443F-47BA-9F96-630107A4BD52@gmail.com> <CAJ-Vmok%2Bpagv3SFHzH1uLasqD6sweXQKZ1_etxWYPNro6vEQZg@mail.gmail.com> <43D75544-1458-4AE1-BCC1-DC5DB5AB72DA@gmail.com> <1428194276.82583.144.camel@freebsd.org> <BF91F228-A5F7-4553-B578-5050A9D73E16@gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 2015-04-04 at 17:41 -0700, Garrett Cooper wrote:
> On Apr 4, 2015, at 17:37, Ian Lepore <ian@freebsd.org> wrote:
> 
> > On Sat, 2015-04-04 at 17:36 -0700, Garrett Cooper wrote:
> >> On Apr 4, 2015, at 17:33, Adrian Chadd <adrian@freebsd.org> wrote:
> >> 
> >>> On 4 April 2015 at 17:25, Garrett Cooper <yaneurabeya@gmail.com> wrote:
> >>>> On Apr 4, 2015, at 17:22, Adrian Chadd <adrian@freebsd.org> wrote:
> >>>> 
> >>>>> On 4 April 2015 at 16:33, Garrett Cooper <yaneurabeya@gmail.com> wrote:
> >>>>>> On Apr 4, 2015, at 16:28, Adrian Chadd <adrian@freebsd.org> wrote:
> >>>>>> 
> >>>>>>> Hi,
> >>>>>>> 
> >>>>>>> cc  -O -pipe   -I/usr/home/adrian/work/freebsd/uverse/head/lib/libcrypt
> >>>>>>> -std=gnu99   -Qunused-arguments
> >>>>>>> -L/home/adrian/work/freebsd/uverse/obj/arm.arm/usr/home/adrian/work/freebsd/uverse/head/tmp/usr/lib/private
> >>>>>>> -rpath /usr/lib/private -rpath /usr/lib/private -o crypt_tests
> >>>>>>> crypt_tests.o /home/adrian/work/freebsd/uverse/obj/arm.arm/usr/home/adrian/work/freebsd/uverse/head/lib/atf/libatf-c/libatf-c.so
> >>>>>>> -lcrypt
> >>>>>>> cc: error: no such file or directory:
> >>>>>>> '/home/adrian/work/freebsd/uverse/obj/arm.arm/usr/home/adrian/work/freebsd/uverse/head/lib/atf/libatf-c/libatf-c.so'
> >>>>>>> 
> >>>>>>> .. this happens for make universe on at least amd64, arm and armeb.
> >>>>>>> I'm still waiting for the others to (not) complete.
> >>>>>> 
> >>>>>> Might be related to r280179 and r273449, but more info´s needed. Your src.conf for the build host and targets would be helpful.
> >>>>>> Thanks!
> >>>>> 
> >>>>> adrian@bruce:~/work/freebsd/uverse/head % cat /etc/make.conf
> >>>>> cat: /etc/make.conf: No such file or directory
> >>>>> adrian@bruce:~/work/freebsd/uverse/head % cat /etc/src.conf
> >>>>> cat: /etc/src.conf: No such file or directory
> >>>>> 
> >>>>> adrian@bruce:~/work/freebsd/uverse/head % env
> >>>>> MAKEOBJDIRPREFIX=/home/adrian/work/freebsd/uverse/obj make -j2
> >>>>> universe JFLAG=-j8
> >>>>> 
> >>>>> Thanks for replying so quickly!
> >>>> 
> >>>> -j2 and JFLAG probably isn´t the best idea in the world. JFLAG should be ok...
> >>> 
> >>> Why not?
> >> 
> >> Based on personal experience, parallelizing on two levels just seems like a bad idea, especially with recursive make and when dealing with .PHONY targets. I haven´t followed through the logic exactly to know where it falls apart, but I know I always run into issues when I use -j instead of JFLAG for specifying -j.
> >> Cheers,
> > 
> > Strange.  I've never heard of JFLAG and have only ever used -j
> 
> JFLAG’s a make tinderbox/universe thing. -j would build multiple universe targets in parallel with up to -j<foo> jobs each.
> 
> Again, I don’t know where and why things fall apart… it’s just that running multiple universes for me -j<foo> tends to get me in trouble. I’d have to go through and audit uses of ${.OBJDIR} to get a better idea of why things fail...
> 
> HTH!

Maybe the j*j thing happened with fmake, I vaguely remember that.  With
bmake the -j value propagates in some magic way so that even though
you're building arches in parallel, the total number of jobs is the -j
value (which you can see pretty clearly in top, after a few minutes the
load averages pretty much lock onto the -j value plus some fraction).
I've never had any problems with objdir stuff building all the arches in
parallel at any value for -j (but I typically never use numbers higher
than 14 unless I'm testing build stuff).

-- Ian





Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1428194986.82583.149.camel>