On 09/12/15 22:28, Karanbir Singh wrote:
> This is a terrible solution. Its a kludge for the best way, and just
> highlights the dire need for EPEL to be better integrated into the
> CentOS side of things.
Actually, I see this as a lack of communication between RHEL and EPEL,
so this problem could have been avoided with better integration between
RHEL and EPEL (not CentOS and EPEL, though of course that would be great
too).
I don't think that the EPEL libunwind maintainer was aware that RHEL was
intending to ship their own libunwind. RHEL libunwind-1.1-5 was
committed in June in order to "beat" EPEL libunwind-1.1-3 (and the RHEL
maintainer told me that their internal review process *does* include a
check for EPEL upgrade path). However, later the EPEL maintainer built
libunwind-1.1-10 and this is where the problems began.
If the EPEL maintainer had been aware of RHEL's libunwind package, he
could have chosen "Release: 3%{?dist}.1" and "Release: 3%{?dist}.2" etc.
for future builds on EPEL to avoid conflicting with RHEL/CentOS. EPEL
wouldn't have needed to retire libunwind so quickly (ie, retire only
after CentOS 7.2 is released), and we wouldn't have had this problem at all.
Kind regards,
Jamie