Many different failures with Fedora-19-Beta-TC3-x86_64-netinst.iso, no successes

[Adam Williamson]

On Sat, 2013-05-04 at 23:18 -0700, Jay Finger wrote:
> I would like to create a bug in bugzilla for this issue, but I can't
> seem to locate a button to do so.  Probably a noob issue.  Regardless,
> how do I do so?

There's only a 'magic button' for reporting crashes - that process runs
through libreport, which is set up to provide relevant info on crashes.
A more generic 'report this failure' button would be possible, but the
infrastructure isn't there yet. I don't know whether libreport would
view this as 'in scope' or not, but we could ask them, I guess.

> Well I should have thought to look in /tmp.  But that also seems like
> a design flaw:  why doesn't anaconda put the log files where people
> know to look for them?

I'm not actually sure, but I think there's a technical reason why they
can't do that; some reason they can't just write to the installer
environment's system directories during installation. The anaconda folks
would be the best to ask.

> 
> In /tmp/packaging.log I see:
> 01:16:01,811 DEBUG packaging: Transaction couldn't start:
> 01:16:01,813 DEBUG packaging: file /usr/bin/ht conflicts between
> attempted installs of ht-2.0.18-4.fc19.x86_x64 and
> texlive-tex4ht-bin-2:svn26509.0-22.20130427_r30134.fc19.x86_64
> 01:16:01.813 ERR packaging:  YumRPMTransError Could not run
> transaction.
> 
> 
> Multiple problems here:
> 1) The error message presented to the user completely sucks.  Showing
> "file /usr/bin/ht conflicts between attempted installs of
> ht-2.0.18-4.fc19.x86_x64 and
> texlive-tex4ht-bin-2:svn26509.0-22.20130427_r30134.fc19.x86_64" would
> not be helpful to most people, but is far better than
> "YumRPMTransError Could not run transaction".

That's true, and it'd be great if you could file an RFE bug for that. It
seems like it ought to be possible.

> 2) The actual useful info is report as a DEBUG message in the log.  

Not sure if that's possible to change, again anaconda folks would know.

> 3) Nothing is presented to the user to report the bug, which means
> that Fedora folks get no feedback as to how many people encounter
> this.

Right, as noted above, we don't have any infrastructure in libreport or
anywhere else for 'intelligently' reporting RPM transaction failures.
It'd require different logic from reporting anaconda crashes. It'd be
nice to have such a thing as well, of course, but then it'd be nice to
have lots of stuff :)

> Having said that, this is _not_ a beta blocker.

Right, as both packages aren't on the DVD.

> I think it should be a release blocker, though.

It ultimately doesn't make sense to block releases on conflicts that are
only present in the remote repos, because we can just as easily fix them
with a package update, so there's no need to go to the trouble of
holding the release of the images.

>   The scenario for this was to select "GNOME Desktop", then select all
> of the additional packages for that environment.

So, it looks like 'ht' is in the Security Lab package group.
texlive-tex4ht-bin is required by texlive-tex4ht, which is in turn
required by:

texlive-collection-htmlxml
texlive-nlctdoc
texlive-scheme-gust
texlive-scheme-xml

texlive-collection-htmlxml is required by dblatex, docbook-utils-pdf,
and texlive-scheme-full; both dblatex and docbook-utils-pdf are in the
"Authoring and Publishing" comps group.

So it looks like, as long as there are these two ht packages, we have a
fundamental conflict between the Security Lab and Authoring and
Publishing group which may not be simple to resolve, unless we can
reconcile the two ht packages. The first, obvious step, though, is to
add an explicit Conflicts: tag to one or both ht packages; this would
cause anaconda to catch the problem much earlier, and with a more useful
error message.

Thanks for reporting!
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net