[Arm-dev] Building failing C7 ARM packages (atlas)

[Fabian Arrotin]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 08/10/15 11:39, PixelDrift.NET Sam wrote:
> On Mon, Oct 5, 2015 at 11:29 PM, Fabian Arrotin <arrfab at centos.org>
> wrote:
>> 
>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>> 
>> On 05/10/15 14:15, PixelDrift.NET Sam wrote:
>>> Hi,
>>> 
>>> I have recently been working through some of the failed CentOS
>>> 7 ARM RPM builds (http://armv7.dev.centos.org/report.html) with
>>> the following successful builds:
>>> 
>>> atlas (F21) 
>>> http://arm.koji.fedoraproject.org/koji/buildinfo?buildID=260230
>>>
>>> 
suitesparse (C7) gegl (C7) scipy (C7) gimp (C7) gsl (C7) bogofilter
>>> (C7) evolution (C7) evolution-ews (C7) evolution-mapi (C7) 
>>> gutenprint (C7) brasero (C7 Modified BuildRequires
>>> brasero-devel) rhythmbox (C7)
>>> 
>>> 
>>> With atlas there appears to be a patch in the Fedora stream for
>>> the same version which fixes the armv7 build issue which
>>> appears to be blocking this package in the automated build
>>> process: * Sun Feb 16 2014 Marcin Juszkiewicz
>>> <mjuszkiewicz at redhat.com <mailto:mjuszkiewicz at redhat.com>> -
>>> 3.10.1-11 - Unbreak AArch64 build. - ARMv8 is different from
>>> ARMv7 so should not be treated as such. Otherwise atlas tries
>>> to do some crazy ARMv764 build and fail.
>>> 
>>> After building atlas with the fixed package I was able to
>>> proceed building the other listed packages with their original
>>> C7 srpms.
>>> 
>>> Can someone advise on the process for getting the patch from
>>> the Fedora package into C7?
>>> 
>>> Thanks in advance.
>>> 
>>> -Sam
>>> 
>> 
>> Well, actually sending the patch (for the .spec and also the
>> patch to be added -if any- as Source) to this list would be
>> welcome. I know that we tried to have it working with the
>> following patch already : 
>> http://armv7.dev.centos.org/repodir/c7-buildroot-patched-SRPMS/patches/atlas-spec.patch
>>
>> 
Not sure if that's the patch used by Fedora, would be good to compare.
>> Let's try to close that, and then release CentOS 7 for arm32,
>> then building all pending updates too. :-)
>> 
> 
> Thanks Fabian,
> 
> The patch looks to be one and the same.
> 
> I downloaded the srpm with the patch applied from here 
> (http://armv7.dev.centos.org/repodir/c7-buildroot-patched-SRPMS/atlas-3.10.1-7.el7.centos.0.1.src.rpm)
>
> 
and it built cleanly on my armv7 hardware (ODROID XU4).
> 
> Is there a build log from the build servers for this patched
> version?

All the build logs are public (and even previous version).
For example , here is the build log for altlas :
http://armv7.dev.centos.org/rpmbuild/c7-pass-1-previous-buildlogs/6695-atlas-3.10.1-7.el7.centos.0.1/armv7hl/build.log

As I'm a sysadmin (read "not a devel guy") and so that infra has
priority #1 for me, I never had a chance to investigate why that one
was failing, nor even tried another time. (I just requeued it though,
let's see)
I don't know if underlying hardware used for the build could make a diff.

The four armv7hl nodes we're actually using have :
cpu model name	: ARMv7 Processor rev 2 (v7l)
Hardware	: Marvell Armada 370/XP (Device Tree)
Revision	: 0000

Those are actually running Fedora 21 with mock-1.1.41-3.fc21.noarch
The idea would be to switch that to c7 as soon as possible too, and
I'm pretty sure that we're close to that, if other people join the "party"

- -- 
Fabian Arrotin
The CentOS Project | http://www.centos.org
gpg key: 56BEC54E | twitter: @arrfab
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)

iEYEARECAAYFAlYXbVQACgkQnVkHo1a+xU6ypQCffF/+hNlYUPU9FdxAMZhVD2pq
9nsAniAAsRi4KkOsznrB0bY9hPSrquLO
=ITz6
-----END PGP SIGNATURE-----