[Buildroot] Analysis of build failures

[Thomas Petazzoni]

Gustavo,

On Wed, 22 Jul 2015 09:51:54 -0300, Gustavo Zacarias wrote:

> >>         sparc |                   boost-1.58.0 | NOK | http://autobuild.buildroot.net/results/dd032aa7b387f3ba5b25bffcaf833ba0ea9ba199/
> >
> > sparc-linux-g++: error: unrecognized argument in option '-mcpu=c3'
> >
> > Gustavo, any idea?
> 
> Hi.
> C3 sound very much like an x86 cpu (Via), i'll take a peek.

Interesting, an x86 -mcpu option used on sparc. Is boost smooking
weed? :-)

> >>       powerpc |            host-heimdal-1.6rc2 | NOK | http://autobuild.buildroot.net/results/a07198b3f7f8029d80863a998814877ec9994c60/
> >>       powerpc |            host-heimdal-1.6rc2 | NOK | http://autobuild.buildroot.net/results/a6d22f80938fa19933940a19664c2354e01a871f/
> >>       aarch64 |            host-heimdal-1.6rc2 | NOK | http://autobuild.buildroot.net/results/cd15555cf90dddacdbf9599313e6e0a118049b38/
> >
> > Parallel build problem. We'll make host-heimdal build with $(MAKE1) in
> > the mean time. Gustavo, can you submit a patch to this effect?
> 
> Ok, seems the sanest until upstream fixes the issues.

I agree.

> >>       powerpc |                 iproute2-4.1.1 | NOK | http://autobuild.buildroot.net/results/5c15e180973425a742629e9e8d0a56e1913d8c03/
> >
> > Smell very much like a too old kernel headers problem. Gustavo?
> 
> Possible, though you know i generally hate those powerpc toolchains :)
> I think it's time to kill them.

I know you hate them. But I like them for one thing: they force us to
keep testing with older kernel headers. Look at Luca Ceresoli, who is
stuck with a 2.6.30 kernel.

> >>       powerpc | libsigrok-fe473123ba836445d... | NOK | http://autobuild.buildroot.net/results/f02e721e1812d03b5990884aa8f8ff09dd88272c/
> >>       powerpc | libsigrok-fe473123ba836445d... | NOK | http://autobuild.buildroot.net/results/bf17b998dcdb81f37776765ca200e0e5c344bada/
> >>       powerpc | libsigrok-fe473123ba836445d... | NOK | http://autobuild.buildroot.net/results/1b38f55602c7aff2afae6f2588e0a8f04b80b09d/
> >>       powerpc | libsigrok-fe473123ba836445d... | NOK | http://autobuild.buildroot.net/results/b7d703cf97345463af373217fb90e076bb3dbeb6/
> >
> > Compiler too old. Maybe exclude the toolchain for now?
> 
> Ditto above, at this point i don't think these will benefit anyone.

Well, C11 is really new. Anyone using gcc 4.7 will hit these problems I
believe, not only these oldish PowerPC toolchains.

> >>           arm |                  netperf-2.6.0 | NOK | http://autobuild.buildroot.net/results/077714a0ee67057abcff680b244228a409243bbf/
> >
> > Smells like a GCC 5.x issue:
> >
> > nettest_omni.o: In function `send_omni_inner':
> > nettest_omni.c:(.text+0x5688): undefined reference to `demo_interval_tick'
> 
> There's a new release, worth trying a bump.

Right. Will you try this?

> >>          i686 |                  openssh-6.9p1 | NOK | http://autobuild.buildroot.net/results/dce0202e039f4636d68532c4aab8738938b76650/
> >
> > Weird build issues. Gustavo?
> 
> Seems like the portability layer is playing games, on a static build, 
> that looks like it's triggering it, i'll investigate.

Great, thanks!

> >>           arm |                  webkit-1.11.5 | NOK | http://autobuild.buildroot.net/results/c7fc621245351eb31afab1b96d3f1f2e6ad0bfad/
> >
> > checking for GLES2/gl2.h... yes
> > checking whether to use OpenGL ES 2 support... configure: error: Cannot enable OpenGL ES 2 support without EGL
> > package/pkg-generic.mk:146: recipe for target '/home/buildroot/build/instance-0/output/build/webkit-1.11.5/.stamp_configured' failed
> > make: *** [/home/buildroot/build/instance-0/output/build/webkit-1.11.5/.stamp_configured] Error 1
> > make: Leaving directory '/home/buildroot/build/instance-0/buildroot'
> >
> > Bernd, Gustavo?
> 
> I think the way forward is upgrading, there's no point in trying to fix 
> such an old version full of security bugs.
> I've got a minimal dep fix as pointed by Yann on the latest version of 
> the webkitgtk24 package, i was waiting on more (any!) feedback from the 
> bunch of users wanting a newer/working webkitgtk/midori but they've been 
> dissapointing to say the least.

Fully agreed. I'm planning on merging your webkit work soonish. It
should definitely be in 2015.08 IMO.

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com