On 19-08-16 23:40, Bjarne Saltbæk wrote:
>
> Hi Gordan.
>
>
> On 19-08-2016 10:49, Gordan Bobic wrote:
>> On 2016-08-18 18:32, Bjarne Saltbæk wrote:
>>> Hi Gordan and Jacco.
>>>
>>>> True, this is an issue, particularly with repository syncing. The
>>>> discrepancy in part comes from the fact that building and signing
>>>> are separate steps in the process
>>>
>>> Here is where Koji have its advantage. My koji installation signs the
>>> package automatically in the end of the build process.
>>
>> Indeed, but my big ARM machine is internet facing, and I am not
>> entirely comfortable keeping the signing keys on a machine that
>> isn't air gapped.
>>
>
> You can do the semi paranoid thing and put two netcards (or just two
> VLAN's on the same netcard) in the Sigul Bridge.
> Then connect one VLAN to the internet, one VLAN to a transport subnet
> to your Sigul Server.
> This is the intended way and this isolates the Sigul Server with the
> pgp signing key from the network.
> The completely paranoid is to shut down /remove the Sigul Server when
> not in use (if you those so I think you should make some email
> notification when there are packages waiting for signing).
>
>
>
>> I'm in the same boat, I haven't had a chance to touch anything
>> RSEL related in weeks. :-(
>>
>> I think there is a lot of value in having an automated system
>> churning out package updates as and when they appear upstream.
>> Even if there are FTBFS packages, if we have a list of those
>> somewhere visible, then whoever from the community needs them
>> can step up and fix them.
>>
>> I've been thinking about this a lot and the more I think about
>> it the more I am leaning toward putting everything on github.
>>
> My experience with Gitblit (which I like very much) is that storing
> bigger files in git makes "git clone" break down (I think something
> with HTTP Chunked not correct implemented in Gitblit).
> Dunno if the same applies for Github.
> I prefer to do the same as Fedoraproject - store .spec file and
> patches in Git, store tarballs outside on a filestore.
>
>>> I have not spent time on EL7 but why not joining forces with the
>>> CentOS7 arm team? Is there any problem with that?
>>
>> It is already happening to a large extent. They took a whole
>> bunch of Jacco's patches to fix various EL7 FTBFS issue on ARM
>> a while back, and since CentOS is effectively our upstream for
>> EL7, we are benefiting from any fixes they apply.
>>
>
> By the way. Isn't CentOS 7 ARM entirely focusing on 64-bit on ARMv8 ?
> If so we could focus on 32-bit RSEL7 (CentOS 7 ARMv6,ARMv7l) ?
> Isn't RSEL7 32-bit compiled? (it runs on ARMv6 and ARMv7l....)

I think there are two distinct CentOS ARM projects:
* one is indeed for ARMv8 (aarch64). They want to standardize on one
kernel (newer then std CentOS) and only support that kernel and UEFI.
https://wiki.centos.org/SpecialInterestGroup/AltArch/AArch64
* one is for ARMv7
https://wiki.centos.org/SpecialInterestGroup/AltArch/Arm32 They are
(like us) building on top of kernels supplied by the vendor of the
specific board.

RSEL7 is ARMv5 compiled and works also on 6 and 7. I guess it'll also
work on 8, but that is yet untested.

Jacco
_______________________________________________
users mailing list
[email protected]
https://lists.redsleeve.org/mailman/listinfo/users

Reply via email to