Bug #17676
Clean up leftovers of src:network-manager 1.6.2-3+deb9u2+0.tails1 in our custom APT repository
100%
Description
Seen while preparing the 4.6 release:
ssh reprepro@incoming.deb.tails.boum.org tails-merge-suite "$RELEASE_BRANCH" "$TAG"
Will not copy as not found: gir1.2-networkmanager-1.0.
Exporting indices...
tails-diff-suites
seems to indicate a match between both suites, so we should be good; but it would be nice to consider the extra clean-up to avoid such noise next time.
Subtasks
History
#1 Updated by CyrilBrulebois 2020-05-04 16:17:36
- blocks Feature #16209: Core work: Foundations Team added
#2 Updated by intrigeri 2020-05-12 13:07:03
- Status changed from Confirmed to Needs Validation
In Bug #17572, we removed the binary gir1.2-networkmanager-1.0
package, but we left a bunch of other obsolete Stretch-era binary packages built from src:network-manager
1.6.2-3+deb9u2+0.tails1:
$ ssh reprepro@incoming.deb.tails.boum.org reprepro list stable | grep -F '1.6.2-3+deb9u2+0.tails1'
stable|main|amd64: libnm-glib-dev 1.6.2-3+deb9u2+0.tails1
stable|main|amd64: libnm-glib-vpn-dev 1.6.2-3+deb9u2+0.tails1
stable|main|amd64: libnm-glib-vpn1 1.6.2-3+deb9u2+0.tails1
stable|main|amd64: libnm-glib-vpn1-dbgsym 1.6.2-3+deb9u2+0.tails1
stable|main|amd64: libnm-glib4 1.6.2-3+deb9u2+0.tails1
stable|main|amd64: libnm-glib4-dbgsym 1.6.2-3+deb9u2+0.tails1
stable|main|amd64: libnm-util-dev 1.6.2-3+deb9u2+0.tails1
stable|main|amd64: libnm-util2 1.6.2-3+deb9u2+0.tails1
stable|main|amd64: libnm-util2-dbgsym 1.6.2-3+deb9u2+0.tails1
+ same in the testing and devel suites.
… so tails-merge-suite
tries to copy all binary packages built from src:network-manager
1.6.2-3+deb9u2+0.tails1, and complains that one of them is missing (because we removed that one).
I’ve verified that none of these packages are listed in tails-amd64-4.6.{build-manifest,packages}
.
I think we should remove these 9 obsolete binary packages from the stable, testing, and devel APT suites.
Once that’s done, we can verify that reprepro dumptracks $DIST
does not mention src:network-manager
1.6.2-3+deb9u2+0.tails1 anymore, and:
- If
src:network-manager
1.6.2-3+deb9u2+0.tails1 is not tracked there anymore, then we’re good: the line of noise this issue is about will disappear. - Else, we may have to research why reprepro thinks that source package is still needed, and possibly go use a lower-level reprepro command to make it forget it.
I’d appreciate if someone (anonym, kibi?) could review this analysis + proposal.
If it makes sense to you, please either reassign to me for implementation… or beat me to it :)
#3 Updated by intrigeri 2020-05-12 13:12:31
- Subject changed from Possible clean-up required in APT repository to Clean up leftovers of src:network-manager 1.6.2-3+deb9u2+0.tails1 in our custom APT repository
#4 Updated by anonym 2020-05-14 10:47:04
- Status changed from Needs Validation to Resolved
- % Done changed from 0 to 100
intrigeri wrote:
> In Bug #17572, we removed the binary gir1.2-networkmanager-1.0
package, but we left a bunch of other obsolete Stretch-era binary packages built from src:network-manager
1.6.2-3+deb9u2+0.tails1:
>
> […]
>
> + same in the testing and devel suites.
>
> … so tails-merge-suite
tries to copy all binary packages built from src:network-manager
1.6.2-3+deb9u2+0.tails1, and complains that one of them is missing (because we removed that one).
> I’ve verified that none of these packages are listed in tails-amd64-4.6.{build-manifest,packages}
.
Yup! Clearly I just missed these packages (presumably I saw the list of libnm*
packages, and saw that the correct versions were at the beginning and end, but ignored the middle :)).
> I think we should remove these 9 obsolete binary packages from the stable, testing, and devel APT suites.
>
> Once that’s done, we can verify that reprepro dumptracks $DIST
does not mention src:network-manager
1.6.2-3+deb9u2+0.tails1 anymore, and:
>
> * If src:network-manager
1.6.2-3+deb9u2+0.tails1 is not tracked there anymore, then we’re good: the line of noise this issue is about will disappear.
> * Else, we may have to research why reprepro thinks that source package is still needed, and possibly go use a lower-level reprepro command to make it forget it.
>
> I’d appreciate if someone (anonym, kibi?) could review this analysis + proposal.
Makes sense!
> If it makes sense to you, please either reassign to me for implementation… or beat me to it :)
Done! The package isn’t tracked any more, so we’re all good! Closing!