Mailinglist Archive: opensuse-project (76 mails)

< Previous Next >
[opensuse-project] Re: OBS (was: Re: 019-04-02 board meeting minutes)

On nie, Apr 21, 2019 at 5:11 PM, Lars Vogdt <lrupp@xxxxxxx> wrote:
On Sun, 21 Apr 2019 01:02:09 +0200 Stasiek Michalski
<hellcp@xxxxxxxxxxxx> wrote:
>> But why? It encourages to contribute to the software which you
>> can run on your favourite distro afterwards :D
>
> No it doesn't - something built in OBS for Fedora or Ubuntu isn't
> magically suddenly available for openSUSE. Other distributions have
> other standards (I would argue lesser ones), and we shouldn't
> compromise openSUSE's quality needlessly.

I mean, that's a fair point, but what if Fedora had an official
instance interconnected with openSUSE instance.

This would - at least for me - impli that Fedora will shut down (or at
least use less) their own build system. If you have any contacts to
developers/decision makers on Fedora site, I'm happy to jump into the
discussion with them.


I feel like the bigger
issue here is that there were trials of making OBS work on other
distros, but were never considered beyond that.

Let's discuss/see if those trials came from openSUSE people, who tried
to convince other distributions from OBS - or real decision makers from
other distributions, who tried to get it to work on their distributions?

Debian jumped onto OBS train because of clang rebuilds back in the days,
although they ultimately decided to keep using their long trusted system of
doing everything on developers machines. The old instance that was made for this
purpose was since abandoned [1]. Trials of porting didn't come "officially" from
distributions, people porting OBS were doing it in their own private time for own
purposes of running OBS on their distro, with Fedora those hopes amounted to
actual code though [2,3], with GPL's hopes of being useful, without any warranty
:P

It's almost as if people doing that hard work
on porting are afraid of the reaction of OBS team to this kind of
submission,
even though it would not hurt openSUSE packaging in any way. There is
a huge
__**BUT**__ here though, it would make SLE11 (general support just
ended, still
on LTSS support for 3 years) incapable of running OBS due to lack of
systemd, and incompatibility of sysv scripts between SLE and RHEL
just makes this harder to support fully.

I guess we have some misunderstanding here.

1) OBS is not an official product

While SUSE is using OBS to build his packages and products, OBS is not
selled to customers as product on its own. So there is no need for SUSE
to do nasty stuff or support OBS to run on old distributions.

JFYI: all OBS instances run either on SLE15 or even SLE15-SP1.

That's a relief to say the least.

The worker images run even sometimes Tumbleweed or the latest Leap
(15.1 in this regard). This depends more on the reliability (kernel,
toolchain) of the distro on different architectures and the needs of
the to build packages than anything else.


2) Support for SysVinit is still there for a reason

We still have some glitches, if the services are managed via systemd.
Especially the "reload" or "restart" handling might be easy for generic
services, but sadly not many of the OBS services are generic.
Just have a look at the OBS sources and issues in Github - we have some
open issues there and working since a while on fixes for them. Feel
free to join the discussions there and drive the porting from systemd
to sysvinit to a success.

*sysvinit to systemd

I will gladly take a look, although assuming that it's related to a bunch
of perl, I will just end up looking at it for hours without doing much
progress.

3) Time - the ugly enemy

If we had enough time, I'm sure that some of my team members whould try
a lot of things. Even to get OBS working on all possible distributions.

But I have to admit that the BuildOPS team (FYI: we are responsible for
package reviews, help with product building, the SUSE maintenance chain,
OBS backend development, PackageHUB, OBS operations and other, non-OBS
related stuff) has not enough manpower to spent this time during
working hours.

So - for me - it is just logical that the SUSE team behind OBS
concentrates on the deliverables they have towards their sponsor SUSE.
That does not mean that we as developers would not accept merge
requests or patches (and indeed, I can show you a lot of patches from
other distributions that went into OBS code) - that just means that our
focus is not on this task.

Then again, I am not expecting SUSE to stop doing everything for a month and
collectively make everything compatible with every distro out there, because it's
unreasonable. Realistically a decision to support another distros should be
decided in early stages of development of any code, but I do not think that
anybody could have expected OBS to prove itself to be as useful as it currently
is.

It kind of is in the hands of the community (not just openSUSE community) to do
something about this though.

I have to admit that I did not see a concrete point in your blaming of
OBS developers for not being supportive enough (I guess you did not
want to blame anyone personally, but please accept that your claims
could be interpreted in this way). But I offer my help (and I'm up
for personal discussions, if you like) here to clarify any
misunderstandings.

Because they are not to blame, historically (open)SUSE has not been excellent at
sharing technologies. At least not in cases where the work is done in
non-standardized environments (like filesystem standards, system inits before
systemd etc). This is obviously complaining about ideology, not about actual
development or developers of the products, because from where I'm standing, they
are doing more than an excellent job.

[1] https://irill8.siege.inria.fr/project/show/Debian:Unstable:Clang
[2] https://github.com/OlegGirko/open-build-service/tree/fedora-2.9.5
[3] https://github.com/strzibny/obs-for-fedora

LCP [Stasiek]
https://lcp.world


--
To unsubscribe, e-mail: opensuse-project+unsubscribe@xxxxxxxxxxxx
To contact the owner, email: opensuse-project+owner@xxxxxxxxxxxx

< Previous Next >
Follow Ups