Broken dependencies in EPEL - 2008-12-07

Michael Schwendt mschwendt at gmail.com
Tue Dec 9 20:52:08 UTC 2008


On Tue, 9 Dec 2008 11:25:11 -0600 (CST), Mike wrote:

> > It had started with a working setup (EL4 and EL5 in a single report):
> > https://www.redhat.com/archives/epel-devel-list/2008-April/msg00016.html
> >
> > Then, sometime later, the RHEL repos have been renamed and moved around
> > again and again (with games like switching repo ids from upper-case to
> > lower-case and vice versa), and finally the shell script has been damaged
> > as an added bonus.
> >
> 
> Actually that only seemed to have worked.  IIRC, we were missing some
> repos so there are false positives there.

Minor problem. Step 1) Make the RHEL repos available. Step 2) Give out
valid (albeit internal) URLs to a person who creates a repoclosure
config script. I've provided the original scripts (list archives -> Feb)
plus even several updates.

Step 1 has been troublesome according to various messages related
to making the same repos available to the buildsys.

Step 2 has never happened. Only with your most recent reply. That is
progress (provided that the repo urls are correct). Only the knowledge of
the full list of repoids makes it possible that someone other than you can
supply a script which contains defaults.

> > I've suggested fixes for the script before based on commit diffs I've seen
> > flying by on sysadmin-members list. It just needs somebody with commit
> > access to fix the stuff or revert to the old working version. And that
> > seems to be the problem, as the RHEL repos are hidden somewhere where it
> > needs special privileges to access them. At present, the RHEL4 URLs used
> > in the script are invalid. For months the report for EPEL4 has been empty
> > because of that. The process is overly complicated. Mike seems to feel
> > pissed as he doesn't like to spend time on it. And stahnma presumably
> > doesn't have the proper permission to fully test and commit his changes.
> > Right?
> >
> 
> Yep, just needs someone with commit access.  Strange that you've never
> offered to do this.

Funny that I've done it for Fedora Extras, spending additional time on
fixing repoclosure to do the things our packagers wanted it to do.

Once some people at the controls started to lock down things, it became
more difficult to impossible. Nowadays we have arrived at a point, where
you want potential contributors to spend hours on IRC waiting for the few
power-hungry "leaders" to move forward and do things where the
unprivileged folk cannot help.

My access has been removed gradually. Earlier this year with the "cleanup"
of sysadmin-build, then later after the break-in with rebuilding of
machines and no documentation of what access is left.

> As we say in the states "Talk is cheap".

Then let me return "Talk is silver, silence is golden". You only flame
while I have had reason to complain. Mind you, I've tried to assist you
with fixes for the commit diffs that are posted to sysadmin members. And
I've also pointed out mistakes and the necessary fixes for the broken
rewrites mstahnma has worked on. That has been fruitless. Weeks have gone
by without progress or requests for help.

It would have taken you only a few minutes to brief me with regard to what
access I can get to roll out a fix myself.

> Bringing
> loose complaints to this list is not even close to the right thing to do
> and while I'd normally let it slide, I'm going to spell it out because you
> should know better, you're an experienced valued contributor.

Valued? I don't have that impression.

> We have a ticketing system.  If you want something done go there.

Ridiculous. As if all tickets would move forward magically.

Like  https://www.redhat.com/archives/fedora-infrastructure-list/2008-October/msg00054.html

Yet nobody else on the list has replied or told me what I could do to
help. I can't even access the epel/extras build machine anymore.

> After a year of this script not working and multiple
> rewrites.

The original shell script was simple. It suffered only from incomplete
repo contents (RHEL server vs. client, and virt.bits) and typos.
The questionable rewrites have not been discussed on any relevant list,
perhaps only on IRC. They've come as a surprise.

> I've taken a calculated decision not to work on it.  It sort of
> works, it emails people, etc.

If you don't care about it, disable it. It mails out crap.
If EPEL has no interest in a broken deps report, discuss that in a
meeting and get rid of it. If there's interest in it, make it possible
that an arbitrary volunteer can take care of it.

> > Rather than spending hours on irc learning about any infrastructure secrets
> > and getting confirmation that some stuff is being locked down,
> > how about posting a verified (!) list of RHEL/EPEL repository URLs?
> 
> And thus the problem, you've become a complainer.

Sure, to shut down one source of false positives. Enough packagers
don't believe broken deps reports. It doesn't help if there is additional
FUD and an installation that mails out crap.

Did you want to continue with these broken deps wreckage till Xmas?

> Think real hard about
> what you're trying to accomplish.  Then ask yourself how the above
> attitude will actually accomplish that.  You've asked for the list (not
> that it will be helpful to you) so I've included it.

It is beyond my comprehension what you try to accomplish with such an
arrogant ending of your mail. You flame me in the "put up or shut up" line
of thinking and at the same time admit that the repo urls are useless to me.




More information about the epel-devel-list mailing list