On Thu, 2016-12-01 at 17:03 -0500, Przemek Klosowski wrote:
> This got me thinking if there's a common root cause that could be
> checked automatically? I didn't quite understand what exactly happened
> in the affected packages to cause it.
No, they're usually all different little awkward packaging corner
cases.
For instance, one common one in the F24 -> F25 upgrade involved the rpm
python subpackages. These were called rpm-python and rpm-python3 in
F24, but in F25 they were renamed to python2-rpm and python3-rpm . Of
course, the F25 packages got lines like:
Obsoletes: rpm-python < %{version}-%{release}
Provides: rpm-python = %{version}-%{release}
Obsoletes: rpm-python3 < %{version}-%{release}
Provides: rpm-python3 = %{version}-%{release}
Unfortunately, the F24 stable 'rpm' package actually got *ahead* of the
F25 stable rpm package for a while. So when you tried to run the
upgrade, the obsoletion didn't kick in - because the F24 package
*wasn't* "< %{version}-%{release}". But dnf couldn't keep the old rpm-
python(3) package(s) around because then some other dependency chain
wasn't satisfied (I forget the details). So it simply had no way to
resolve the problem without removing everything that required rpm-
python or rpm-python3 ...
But that's just one possible case, there have been many others.
Packaging is hard. You can usually figure it out, if you dig into a
bit; it *does* help to file bugs so the issues can be solved for
others.
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net
_______________________________________________
devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx
