[Buildroot] [autobuild.buildroot.net] Build results for 2016-05-13

[Thomas Petazzoni <thomas.petazzoni@free-electrons.com>]

Hello,

On Sat, 14 May 2016 12:02:11 -0300, Gustavo Zacarias wrote:

> > Gustavo, this is the same issue as the one we add with the old
> > CodeSourcery toolchain, but this time with a Crosstool-NG toolchain
> > that uses gcc 4.7. What is so special with blktrace that it causes such
> > issues?  
> 
> This toolchain must also die, since it's glibc-based we can assume it's 
> bad for SPE (likely eglibc). And blktrace needs glibc/musl, hence for 
> SPE it's a no go in the end (i believe Rich was looking into SPE support 
> for musl, but will probably go for soft-float given the kind of broken FPU).

I still don't understand all this PowerPC/ABI/FPU/libc stuff. I guess
it would be useful to write down once for all on the Buildroot Wiki
what is the situation in this area.

> That being said, i think this is an old gcc bug, however i'm failing to 
> find the exact details.

What I find weird with this issue is that only blktrace seems to
trigger the issue. Is blktrace doing something funky? Most other
packages appear to build fine with this toolchain.

> > cairo-gl.h:130:21: fatal error: EGL/egl.h: No such file or directory
> >   #include <EGL/egl.h>
> >
> > Gustavo, Bernd ?  
> 
> It's mali-t76x's fault, since the package only installs binaries it 
> depends on mesa3d-headers for the headers, but in fact it doesn't, so 
> cairo builds before it.
> Patch sent.

I've applied the patch.

> >> microblazeel |                 collectd-5.5.1 | NOK | http://autobuild.buildroot.net/results/602f1b40d1c369bb7eef24216264e0e41837518e/
> >>           arm |                 collectd-5.5.1 | NOK | http://autobuild.buildroot.net/results/bf6e4d55b3ff069c8efd72721a65af2c30e5b6f4/  
> >
> > There's a dependency error on iptables. Gustavo, could you have a look?  
> 
> This is the big iptables/netfilter header debacle from 4.5.x, i can't 
> stress how much this thing will linger into future breakage.
> It will need some preprocessor hacky fix, or disabling iptables support 
> which wouldn't be too nice (even if it's for 4.5).

Are you going to work on this topic?

> >>          i686 |                 libinput-1.2.4 | NOK | http://autobuild.buildroot.net/results/3eb32c19f90a5fd8d45a0c36676e015e8278a469/  
> >
> >    CCLD     event-debug
> > ../src/.libs/libinput.so: undefined reference to `static_assert'
> >
> > Baruch, you recently looked at libinput, could you investigate this
> > build failure?  
> 
> static_assert() needs "modern" C11 support, and the feature isn't 
> checked for in configure, that's the problem.
> Quickie fix could be an #ifdef since it's basically a build check.

Baruch has sent a fix, I'll apply it.

> >>         sparc |                     mpv-0.17.0 | NOK | http://autobuild.buildroot.net/results/c77e763817657b251213a7ce1e52513b3233e9cd/
> >>           arm |                     mpv-0.17.0 | NOK | http://autobuild.buildroot.net/results/88c34347cb4d0f33221415f00426be3dc4223d3b/
> >>       powerpc |                     mpv-0.17.0 | NOK | http://autobuild.buildroot.net/results/cc4b52b5a222cad30c369be493d918916df3ec70/
> >>      mips64el |                     mpv-0.17.0 | NOK | http://autobuild.buildroot.net/results/011c74cce1ca77fe4c7880226f43ef6e5bc6aadc/
> >>           arm |                     mpv-0.17.0 | NOK | http://autobuild.buildroot.net/results/663da681c22a3c913de174b6b80341e2f8e6ca33/
> >>         sparc |                     mpv-0.17.0 | NOK | http://autobuild.buildroot.net/results/74787def01b321f7b9c6736c6871f8558d23b1ab/  
> >
> > Gustavo, you added mpv recently, could you look into those issues?  
> 
> Most of these are lack of xsi math in the buildroot toolchains which 
> need to get updated.

OK, I'll rebuild the Buildroot toolchains.

> The only exception is sparc that lacks atomics and thus needs a fix.

OK. Will you send a patch for this?

> >>          bfin |                     ruby-2.3.1 | NOK | http://autobuild.buildroot.net/results/b4e469a72a746c05ace172798771cfcd081ea473/  
> >
> > process.c: In function 'rb_spawn_process':
> > process.c:3928: warning: passing argument 2 of 'rb_ary_new_from_values' from incompatible pointer type
> > process.c:3930: error: 'status' undeclared (first use in this function)
> > process.c:3930: error: (Each undeclared identifier is reported only once
> > process.c:3930: error: for each function it appears in.)
> >
> > Gustavo? This seems like a recent regression, I remember being able to
> > build ruby on bfin some time ago.  
> 
> It seems so, i'll take a look since it's supposed to work for nommu.

OK, thanks!

> >>          bfin |                 sg3_utils-1.42 | NOK | http://autobuild.buildroot.net/results/31447787e70cd351ce01b7a49ba029758d0c68e5/  
> >
> > Static linking or weird bfin prefix issue. If it's Blackfin related, I
> > would suggest to make this package unavailable on Blackfin.  
> 
> Yes, bfin with SAS/SATA storage is pretty unlikely, so blacklisted.
> It's the usual weird symbol prefixing prefixing problem for bfin flat, 
> so if anyone is inclined to fix then it's welcome, though it's an 
> unlikely scenario since blackfin lacks storage ports compatible with 
> sg3_utils (sas, sata, scsi).

I've applied your patch, thanks!

> >>      mips64el |                   stunnel-5.31 | NOK | http://autobuild.buildroot.net/results/293b9ff94c5e28802f3112eb764b62226aee8bea/
> >>      mips64el |                   stunnel-5.31 | NOK | http://autobuild.buildroot.net/results/b940f2faa567229dd9fd30041430ed3eea96e297/  
> >
> > Vicente, Gustavo, I am not sure to whom of you this falls. It doesn't
> > seem to be architecture related, but only happened on mips64el. Could
> > you have a look?  
> 
> It's consistent for certain combinations of mips, so i defer to Vicente 
> (i believe again!) on this.
> Always R6 ISA, little endian, external toolchain it seems.

Vicente? :-)

Thanks,

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