I've adjusted the script to work around the qemu sorting and report a bit more info.
all here: https://mikem.fedorapeople.org/
(script has been replaced, the current log is centos_srpm_sort2c.log.

One of the changes is that I had it report when package names are completely missing from one side or the other (as opposed to just missing a build somewhere).  This leads to more oddities

5) vdsm-4.9-63.el6.src.rpm

Centos has just one build of vdsm, which does not appear on ftp.rh.c. The rh ftp only has vdsm sources under RHEV and RHS, and nothing quite so old as 4.9-63.  Does anyone know the story behind this build? Was it pulled in for a reason, or is this just a case of an accidental ftp posting that got rebuilt before someone noticed and removed it?

6) a number of package names unique to rhel

Many of these make sense. Heck, they probably all have good reasons, but I figure as long as we're here it would be nice for folks to have a look and make sure. Most seem to fall into the following categories:
- rhn-related stuff (e.g. subscription-manager)
- rh release/branding related stuff (e.g. redhat-release-*)
- arch specific stuff (e.g. s390utils)
- eal4 cert related (cc-eal4-config-rhel62)

A few seem to related to specialized hardware. I'm not 100% sure why that excludes them from CentOS, but here is that list:
hpwdt
libibverbs-rocee
libmlx4-rocee

Finally there are the ones I'm really not sure about. I suspect that all of these have a good reason for exclusion. I suspect in most cases it is because they are specific to one of the more specialized variants of rhel (e.g. sap, sap-hana)
compat-gcc-295
compat-sap-c++
qpid-qmf
sapconf
snd-hda
vhostmd



On Fri, Nov 14, 2014 at 12:03 PM, Mike McLean <mikem@imponderable.org> wrote:
Some oddities in the data

1) Some odd rebuilds

rebuilds of RHEL6.0 beta packages:
fence-virt-0.2.1-3.el6.src.rpm
libvpd-2.1.1-2.el6.src.rpm

A rhev build thrown in?
vdsm-4.9-63.el6.src.rpm

A number of non-kernel same-nvr rebuilds:
subversion-1.6.11-7.el6.src.rpm
dhcp-4.1.1-34.P1.el6_4.1.src.rpm
zsh-4.3.10-7.el6.src.rpm
zsh-4.3.10-8.el6_5.src.rpm
zsh-4.3.10-9.el6.src.rpm

2) Some kernels seemingly missing rebuilds

kernel-2.6.32-358.46.1.el6.src.rpm
kernel-2.6.32-358.46.2.el6.src.rpm
kernel-2.6.32-358.48.1.el6.src.rpm
kernel-2.6.32-358.49.1.el6.src.rpm
kernel-2.6.32-431.37.1.el6.src.rpm

Not a big deal if they weren't rebuilt. I just want to make sure I'm not
missing anything.


3) A number of intermittent customizations

I suspect these are all normal cases of needing to tweak something one
time to get a build through or to solve a qa issue. Again, just want to
make sure I'm not missing something

ipa
librsvg2
openscap
openssl
pango
qemu-kvm (* see 4)
subversion
virt-who


4) qemu-kvm ordering

I see 5 custom rebuilds of qemu-kvm, but their n-v-r-s all sort lower
than they should. E.g.

qemu-kvm-0.12.1.2-2.355.0.1.el6.centos.2.src.rpm
qemu-kvm-0.12.1.2-2.355.0.1.el6.centos.3.src.rpm
qemu-kvm-0.12.1.2-2.355.0.1.el6.centos.5.src.rpm
qemu-kvm-0.12.1.2-2.355.0.1.el6.centos.6.src.rpm
qemu-kvm-0.12.1.2-2.355.0.1.el6_4.9.src.rpm

all sort lower than:
qemu-kvm-0.12.1.2-2.355.el6_4.2.src.rpm
qemu-kvm-0.12.1.2-2.355.el6_4.3.src.rpm
qemu-kvm-0.12.1.2-2.355.el6_4.5.src.rpm
qemu-kvm-0.12.1.2-2.355.el6_4.6.src.rpm
qemu-kvm-0.12.1.2-2.355.el6_4.9.src.rpm

At this point I'm assuming that the first set are all rebuilds of the
second and that I should manually reorder these for the import (or add
some very special case hacks to the script).



On Thu, Nov 13, 2014 at 12:31 PM, Mike McLean <mikem@imponderable.org> wrote:
I've generated my own list of ordered sources. Following are the techniques, scripts, and data. Comments or corrections on any of these would be most appreciated.

Sources of Data
1) ftp.redhat.com
    all *.src.rpm files under redhat/linux/enterprise/6*/en/os/SRPMS
2) vault.centos.org
    all *.src.rpm files under 6.*/{os,updates,fasttrack}

If anyone feels I'm starting from the wrong data, please say so. I debated about including the fasttrack dirs, but as this point I've convinced myself that it is probably correct, or at worst harmless).

I wrote a script to do all the heavy lifting. For args it expects files containing lists of paths to source rpms (there are too many srpms involved to pass them directly on the command line). The script identifies srpms with identical contents by comparing the list of files and their hashs (so a centos rebuilt srpm with no changes is considered a duplicate of the rh one even though some of the headers change (e.g. vendor, buildtime).

The script sorts first by package name, then by version-release *with dist tag removed*, then by rh vs centos, then by full version-release. An srpm is considered a centos rpm if either the vendor is centos, or a centos dist tag appears in the release, otherwise it is considered an rh srpm.




On Thu, Oct 30, 2014 at 1:06 PM, Johnny Hughes <johnny@centos.org> wrote:
On 10/30/2014 11:50 AM, Mike McLean wrote:
> On Tue, Oct 7, 2014 at 3:42 PM, Johnny Hughes <johnny@centos.org
> <mailto:johnny@centos.org>> wrote:
>
>     As was discussed before, CentOS-6 SRPMS are going to be imported into
>     git.centos.org <http://git.centos.org> as well and will be processed
>     like CentOS-7 ones are now.
>
>     We want to bring everything in from 6.0 initial and through 6.4+updates
>     initially, then we will do 6.5+updates (as that is changing right now).
>
>     So, I have created a 6.0 to 6.4 set of lists.  These lists live at this
>     location:
>
>     http://people.centos.org/hughesjr/EL6-Import/
>
>     The two lists so far are:
>
>     EL6-non-mod-SRPMS-sorted.txt
>     centos-6-srpms-modified.txt
>
>     1.  The EL6-non-mod-SRPMS-sorted.txt is all SRPMS used in CentOS-6 in
>     their unmodified form.  The order they appear in the file is the order
>     they will be imported into git.  What is important for history is that
>     (for each NAME) they are imported in the correct order, so from 6.0
>     through 6.4+updates, the order of packages used in CentOS-6 for
>     389-ds-base would be:
>
>     389-ds-base-1.2.8.2-1.el6.src.rpm
>     389-ds-base-1.2.8.2-1.el6_1.3.src.rpm
>     389-ds-base-1.2.9.14-1.el6.src.rpm
>     389-ds-base-1.2.9.14-1.el6_2.2.src.rpm
>     389-ds-base-1.2.10.2-15.el6.src.rpm
>     389-ds-base-1.2.10.2-18.el6_3.src.rpm
>     389-ds-base-1.2.10.2-20.el6_3.src.rpm
>     389-ds-base-1.2.11.15-11.el6.src.rpm
>     389-ds-base-1.2.11.15-12.el6_4.src.rpm
>     389-ds-base-1.2.11.15-14.el6_4.src.rpm
>     389-ds-base-1.2.11.15-20.el6_4.src.rpm
>     389-ds-base-1.2.11.15-22.el6_4.src.rpm
>
>     All of these packages will come from ftp.redhat.com
>     <http://ftp.redhat.com> and be imported.
>
>
> I'm running a test import of the nonmod ones and it appears a number in
> the list have the wrong dist tag in the name.
> For example, bnx2-2.2.1.32.269-1.el6.src.rpm should be
> bnx2-2.2.1.32.269-1.el6_2.src.rpm (e.g. .el6_2 instead of .el6). Can you
> confirm?

Yes, those were wrong in centos .. for those, I think we have the centos
name.





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