On 13 March 2015 at 13:25, Toni Spets <toni.spets@iki.fi> wrote:
Hello list,

I've been in contact with Anders F Björklund (afb) earlier this week regarding the state of Xfce on CentOS 7 and in general in any supported EPEL release.

What we discussed was what he has done in the past for his personal use and what are the things we would both agree on that should get some attention to make the experience better.

Here are some initial things that we think are critical (give or take, Anders has his own list):

1. Packaging issues
  - comp group @xfce-desktop doesn't depend on @X Window System so it doesn't really work and has some wrong packages and some missing
  - @xfce-desktop depends on gdm instead of lightdm, we think it should do that instead
  - dejavu sans fonts are not a dependency and fonts are broken because of that
  - gtk-xfce-engine package is missing (EPEL7)

2. Branding issues
  - "Default" appearance theme is missing (requires gtk-xfce-engine, request sent to EPEL)
  - default background is blank (broken Fedora branding, patch submitted)
  - default icon set is wrong and broken (should be GNOME because "Fedora" doesn't exist on RHEL, patch submitted)


There are more, but these are one of the first usability issues that you would hit when installing Xfce from EPEL 7.

I have submitted bug reports with proposed fixes for the background issue, default theming issue and missing gtk-xfce-engine package:

https://bugzilla.redhat.com/show_bug.cgi?id=1200988
https://bugzilla.redhat.com/show_bug.cgi?id=1200992
https://bugzilla.redhat.com/show_bug.cgi?id=1201124

Now, getting EPEL fixed up would be in general a good effort since it's shared between all derivatives of RHEL/CentOS and more people will get the base benefits. That said, there are two ways this SIG could be constructed:

a) shim wrapper around EPEL to create a spin and possibly add some CentOS specific branding, give EPEL maintainer all the help we can to get the packages working, try push most if not all packaging work to EPEL
b) completely replace EPEL by our own repo which has all the Xfce packages and anything we would want

The A option depends on if the maintainer is willing to let us help as much as we want and would accept some of the changes we'd need. It would also imply the current EPEL maintainer(s) would continue maintaining and be our upstream or someone would join the EPEL folks as co-maintainers. The amount of initial work would be minimal compared to doing everything custom and as I said before, it benefits every RHEL derivative, including RHEL itself. We still need some SIG specific packages like CentOS specific branding and possibly our own default panel layout so it's not all upstream work. This approach might also be tad slower because we have an upstream and all packaging stuff except barnding would go through them.

The B option gives us more power and could speed up the process a lot. We can select and backport the set of Xfce packages as fast as we can.Karanbin Singh was kind enough to provide some info about how CentOS SIGs work and what resources would be available (automatic builds, official SIG repos etc.) so all the backend stuff is there waiting to be untilized for this kind of SIG. However, this approach would be very CentOS specific and would require the SIG to be fairly active in maintaining the packages when our upstream (the Xfce project) makes point releases. I can say upfront that person wouldn't be me, so if someone is up to this after the initial sprint of work has ended, that would be something to consider.

One more very important thing to note is that EPEL 7 is set to Xfce 4.10 and 4.12 just came fresh out of the oven. If we do A, I'm fairly sure we would need to stick with 4.10 unless the maintainer feels 4.12 upgrade would be possible, then it's a win for option A. The maintainer (or one of them) also maintains a 4.12 copr for F20 so I strongly believe he knows much more about Xfce packaging than me for one. If we do option B we could scrap 4.10 immediately and package 4.12. There is no reason to use 4.10 as a SIG would be the perfect place to get more up-to-date experience on top of stable core.


You would also be repeating work from various other groups who are doing things in either COPRs or the Fedora XFCE mailing list. Kevin Fenzi (irc:nirik) and Adam Miller (maxamillion) are doing a lot of this work in their spare time. Helping them out would move the following issues a lot.

1) A lot of plugins that were in 4.10 and earlier are no longer in 4.12
2) A lot of 4.12 was actually built around a gtk3 and tools which aren't in EL-7. 

Both of these are actually going to take a lot of work with upstream and the software itself to

a) make new plugins which work with the 4.12 framework
b) work on patches or newer upstream releases to make it work with EL-7. 

Once those are done there looks to be plans to get 4.12 in EL-7. I am hoping to get some time to look at scl's for EL-5 and EL-6 and those would be ones which might do fine with a refresh. 

 
I'm up for either one, it's up to what you other people would see as being more useful. Maybe both if there is enough interest? Fix EPEL for the greater good and still have our own SIG 4.12 stuff on top of it for CentOS only.

Anyone into this except me and Anders? Ideas, suggestions?

--
Toni Spets

_______________________________________________
CentOS-devel mailing list
CentOS-devel@centos.org
http://lists.centos.org/mailman/listinfo/centos-devel




--
Stephen J Smoogen.