Hello fellow packagers,
Yep, it's that time of the release cycle again. What should be do about all the packages that are stuck at older versions in f38 than what's available on f37? If you see your package in the list below (with my description of what seems to be wrong with it), please ensure that builds + updates for Fedora 38 are filed soon. Otherwise the update might not make it to stable before the final freeze and will be forever stuck at the old version in the "fedora" repository.
Full list of *version downgrades only* attached below (downgrades considering the full NVR are a much longer list, but the signal-to-noise ratio is also much worse).
One problem that I haven't seen before is that updates are apparently in "testing" or "testing->stable" state in bodhi, but aren't available from the "updates-testing" repository at all. No wonder these updates aren't getting any karma ...
I'm also worried about the unusually large number of builds that fell into the mass branching void this time round (i.e. tagged only with "f39" but not with "f38" in koji, but the build still has the fc38 dist tag).
Fabio
======================================================================
- Lmod-0:8.7.19-1.fc37 > Lmod-0:8.7.18-1.fc38
There is an update for Lmod v0.8.19 in "testing->stable" state, but it's not available from the updates-testing repos.
- autorandr-0:1.13.3-1.fc37 > autorandr-0:1.12.1-3.fc38
This package was only updated to v1.13.3 on the f37 branch, but not on f38 or rawhide branches.
- buildah-0:1.29.1-1.fc37 > buildah-0:1.29.0-2.fc38
Another update that is in "testing" state, but not available from the updates-testing repository.
- clementine-0:1.4.0-2.rc1.20220107git0be3143.fc36 > clementine-0:1.4.0~rc2-3.fc38
Package downgrade due to incompatible versioning scheme change.
- cockpit-composer-0:44-1.fc37 > cockpit-composer-0:43-1.fc38
Packit ignored the f38 branch entirely.
- dcfldd-0:1.9-1.fc37 > dcfldd-0:1.8-2.fc38
Package was updated to v1.9 for rawhide/f39, f37, f36, epel9, and epel8 branches, but f38 was missed.
- dotnet6.0-0:6.0.114-1.fc37 > dotnet6.0-0:6.0.113-2.fc38
The update to v6.0.114 is in "testing->stable" state, but not available from the repositories.
- dotnet7.0-0:7.0.103-1.fc37 > dotnet7.0-0:7.0.102-2.fc38
Same problem. The update for v7.0.103 is in "testing->stable" state, but not available from updates-testing repository.
- enchant2-0:2.3.4-1.fc37 > enchant2-0:2.3.3-6.fc38
This update was probably filed during the signing outage / beta freeze enablement and was stuck in "pending" state. I have just pushed the button to move it to "pending->testing" ..
- freedroidrpg-0:1.0rc2-0.2.fc37 > freedroidrpg-0:1.0-1.fc38
Another package downgrade caused by using a wrong versioning scheme. It's confusing to users who upgrade, but not an actual problem.
- freerdp-2:2.10.0-1.fc37 > freerdp-2:2.9.0-3.fc38
Another two-week-old update stuck in "pending" state. I pushed the "push to testing" button.
- gh-0:2.23.0-1.fc37 > gh-0:2.22.1-1.fc38
Looks like v2.22.1 was tagged into f38 *after* v2.23.0 was tagged, so the older version took precedence ... gotta love koji not knowing how to sort RPM versions :)
- glab-0:1.25.3-1.fc37 > glab-0:1.24.1-1.fc38
Similar problem here. The v1.25.3 update for Fedora 38 was only tagged into f39 but not f38, but has an fc38 dist tag ... probably the build happened in the mass branching void :(
- golang-github-google-dap-0:0.7.0-1.fc37 > golang-github-google-dap-0:0.6.0-8.fc38
Version 0.7.0 is apparently a security update (CVE-2022-41717) but was only built for f37 and f36, but not for f38 or rawhide.
- golang-github-rabbitmq-amqp091-0:1.7.0-1.fc37 > golang-github-rabbitmq-amqp091-0:1.6.1-1.fc38
Version 1.7.0 was built for f39, f37, and f36, but not f38.
- gpodder-0:3.11.1-1.fc37 > gpodder-0:3.11.0-3.fc38
Version 3.11.1 was built for f39, f37, and f36, but not f38.
- hdparm-0:9.65-1.fc37 > hdparm-0:9.63-2.fc38
Another update that was stuck in "pending". I pushed "the button".
- indistarter-0:2.3.1^2022041052da97e-2.fc37 > indistarter-0:2.3.1^202301251abc907-1.fc38
This looks like a harmless post-release-snapshot versioning syntax mistake.
- inxi-0:3.3.25-1.fc37 > inxi-0:3.3.24-2.fc38
Another build with an fc38 dist tag that was only tagged into f39. Mass branching void casualty.
- libstoragemgmt-0:1.9.7-1.fc37 > libstoragemgmt-0:1.9.6-3.fc38
Version 1.9.7 was built for f39, f37, and f36, but not f38.
- mame-0:0.252-3.fc37 > mame-0:0.251-1.fc38
Looks like this downgrade is caused by an F38+FTBFS issue.
- mock-core-configs-0:38.2-1.fc37 > mock-core-configs-0:38.1-1.fc38
Version 38.2 was built for f39, f37, f36, epel9, epel8, epel7, but not f38.
- obs-build-0:20230228-415.1.1.fc37 > obs-build-0:20230208-410.1.1.fc38
The 415.1.1 update was built for f38 but never submitted to bodhi. I pushed the necessary buttons.
- perl-CPAN-Perl-Releases-0:5.20230220-1.fc37 > perl-CPAN-Perl-Releases-0:5.20230120-1.fc38 - perl-Module-CoreList-1:5.20230220-1.fc37 > perl-Module-CoreList-1:5.20230120-1.fc38
Both updates were stuck in "pending". I pushed the "push to testing" button.
- php-laminas-i18n-0:2.21.0-1.fc37 > php-laminas-i18n-0:2.19.0-1.fc38 - php-laminas-navigation-0:2.18.1-1.fc37 > php-laminas-navigation-0:2.16.0-2.fc38 - php-laminas-session-0:2.16.0-1.fc37 > php-laminas-session-0:2.13.0-2.fc38 - php-laminas-validator-0:2.28.0-1.fc37 > php-laminas-validator-0:2.26.0-1.fc38 - php-laminas-view-0:2.27.0-1.fc37 > php-laminas-view-0:2.25.2-2.fc38 - php-symfony4-0:4.4.50-1.fc37 > php-symfony4-0:4.4.47-1.fc38
These appear to be caused by F38+FTBFS issues.
- python-apprise-0:1.3.0-1.fc37 > python-apprise-0:1.2.1-2.fc38
Version 1.3.0 was built for f39, f37, f36, epel9, and epel8, but not f38.
- python-dask-0:2023.2.0-1.fc37~bootstrap > python-dask-0:2023.1.0-1.fc38~bootstrap
Appears to be caused by F38+FTBFS. No idea why the package is in eternal bootstrap hell.
- python-pefile-0:2023.2.7-1.fc37 > python-pefile-0:2022.5.30-3.fc38
Ah, another build that fell into the mass branching void. python-pefile-2023.2.7-1.fc38 has the fc38 dist tag but wound up getting tagged only with f39 in koji.
- python-pynamodb-0:5.4.1-1.fc37 > python-pynamodb-0:5.4.0-1.fc38
Was built for f39, f38, f37, f36, and epel9, but bodhi update was not submitted for f38. I pushed the buttons.
- python-pytenable-0:1.4.12-2.fc37 > python-pytenable-0:1.4.11-2.fc38
Version 1.4.12 was only pushed to the f37 branch, but no others. :shrug:
- python-specfile-0:0.14.0-1.fc37 > python-specfile-0:0.13.2-1.fc38
Another packit victim. The v0.14.0 update is missing entirely from f38.
- python-sport-activities-features-0:0.3.10-1.fc37 > python-sport-activities-features-0:0.3.8-2.fc38
Build appears to have failed only on f38, but succeeded on all other branches (including rawhide).
- python-toml-adapt-0:0.2.11-1.fc37 > python-toml-adapt-0:0.2.10-2.fc38
The update to v0.2.11 was pushed to all branches except f38 (not even in dist-git).
- python-tqdm-0:4.65.0-2.fc37 > python-tqdm-0:4.64.1-2.fc38
Package was built across all branches but the bodhi update for f38 was missing. I pushed the right buttons in the right order to fix that.
- python-uvloop-0:0.17.0-3.fc37 > python-uvloop-0:0.15.3-6.fc38
Version 0.17.0 was only pushed to f37 and f39/rawhide, but not f38.
- python-xlsxwriter-0:3.0.8-1.fc37 > python-xlsxwriter-0:3.0.7-1.fc38
Version 3.0.8 was pushed to f39, f37, f36, and epel9, but not f38.
- rarian-0:0.8.2-1.fc37 > rarian-0:0.8.1-44.fc38
Version 0.8.2 was pushed to f39, f37, and f36, but not f38.
- scrot-0:1.8.1-1.fc37 > scrot-0:1.7-4.fc38
Version 1.8.1 was only pushed to f37, but not f38 or rawhide (even dist-git has only changes on the f37 branch ... maybe packager is confused? This affects a few of their packages).
- sddm-0:0.19.0^git20230306.572b128-1.fc37 > sddm-0:0.19.0^git20230214.8f1e3df-1.fc38
The missing update was built in koji, but not submitted to bodhi. I was not brave enough to push the buttons for this one though.
- syncthing-0:1.23.1-1.fc37 > syncthing-0:1.22.0-1.fc38
Caused by F38+FTBFS.
- toot-0:0.35.0-1.fc37 > toot-0:0.34.1-2.fc38
Version 0.35.0 was built for f39, f37, and f36, but not f38. This is starting to make me sad.
- translate-shell-0:0.9.7.1-1.fc37 > translate-shell-0:0.9.7-2.fc38
Another build that fell into the void during mass branching: translate-shell-0.9.7.1-1.fc38 is only tagged with f39 but still has the fc38 dist tag.
- vkBasalt-0:0.3.2.9-1.fc37 > vkBasalt-0:0.3.2.8-1.fc38
Downgrade appears to be caused by F38+FTBFS.
- wesnoth-0:1.17.13-1.fc37 > wesnoth-0:1.17.7-2.fc38
Update to v1.17.13 was built for f38, but not submitted to bodhi. I pushed the buttons to make that happen.
- yoshimi-0:2.2.3-1.fc37 > yoshimi-0:2.2.1-2.fc38
Version 2.2.3 was pushed to f39, f37, and f36, but not f38.
On Thu, 2023-03-09 at 01:30 +0100, Fabio Valentini wrote:
- gh-0:2.23.0-1.fc37 > gh-0:2.22.1-1.fc38
Looks like v2.22.1 was tagged into f38 *after* v2.23.0 was tagged, so the older version took precedence ... gotta love koji not knowing how to sort RPM versions :)
It's not that it doesn't know, more just that it intentionally works this way. Assuming that we always want the highest NVR ever built for a tag would cause different problems, not fewer problems...
On 09. 03. 23 1:30, Fabio Valentini wrote:
- python-dask-0:2023.2.0-1.fc37~bootstrap >
python-dask-0:2023.1.0-1.fc38~bootstrap
Appears to be caused by F38+FTBFS. No idea why the package is in eternal bootstrap hell.
See https://src.fedoraproject.org/rpms/python-dask/pull-request/4
One problem that I haven't seen before is that updates are apparently in "testing" or "testing->stable" state in bodhi, but aren't available from the "updates-testing" repository at all. No wonder these updates aren't getting any karma ...
hmm that is strange because the builds are in the compose https://kojipkgs.fedoraproject.org/compose/updates/Fedora-38-updates-testing...
On Thu, Mar 9, 2023 at 12:38 PM Tomas Hrcka thrcka@redhat.com wrote:
One problem that I haven't seen before is that updates are apparently in "testing" or "testing->stable" state in bodhi, but aren't available from the "updates-testing" repository at all. No wonder these updates aren't getting any karma ...
hmm that is strange because the builds are in the compose https://kojipkgs.fedoraproject.org/compose/updates/Fedora-38-updates-testing...
Uhm, are you sure? Neither "Lmod", "buildah", nor "dotnet" do not appear in this list at all.
Fabio
On Thu, Mar 9, 2023 at 2:44 PM Fabio Valentini decathorpe@gmail.com wrote:
On Thu, Mar 9, 2023 at 12:38 PM Tomas Hrcka thrcka@redhat.com wrote:
One problem that I haven't seen before is that updates are apparently in "testing" or "testing->stable" state in bodhi, but aren't available from the "updates-testing" repository at all. No wonder these updates aren't getting any karma ...
hmm that is strange because the builds are in the compose https://kojipkgs.fedoraproject.org/compose/updates/Fedora-38-updates-testing...
Uhm, are you sure? Neither "Lmod", "buildah", nor "dotnet" do not appear in this list at all.
Looking into it further, it appears that these builds are indeed mis-tagged in koji (i.e. with "f38-updates-testing-pending" instead of "f38-updates-testing"), while bodhi believes they are in "testing" state (which they are not). I filed a ticket: https://pagure.io/releng/issue/11326
Fabio
Uhm, are you sure? Neither "Lmod", "buildah", nor "dotnet" do not appear in this list at all.
Looking into it further, it appears that these builds are indeed mis-tagged in koji (i.e. with "f38-updates-testing-pending" instead of "f38-updates-testing"), while bodhi believes they are in "testing" state (which they are not). I filed a ticket: https://pagure.io/releng/issue/11326
I checked randomly from the updates list for testing. Thanks for the ticket I will look into it.
- perl-CPAN-Perl-Releases-0:5.20230220-1.fc37 >
perl-CPAN-Perl-Releases-0:5.20230120-1.fc38
- perl-Module-CoreList-1:5.20230220-1.fc37 >
perl-Module-CoreList-1:5.20230120-1.fc38
Newer builds are in testing, because there is note 'Frozen release - This update will not be pushed to stable until freeze is lifted from Fedora 38.'
On Thu, Mar 9, 2023 at 3:25 PM Jitka Plesnikova jplesnik@redhat.com wrote:
- perl-CPAN-Perl-Releases-0:5.20230220-1.fc37 >
perl-CPAN-Perl-Releases-0:5.20230120-1.fc38
- perl-Module-CoreList-1:5.20230220-1.fc37 >
perl-Module-CoreList-1:5.20230120-1.fc38
Newer builds are in testing, because there is note 'Frozen release - This update will not be pushed to stable until freeze is lifted from Fedora 38.'
The problem was that until I pushed the "push to testing" button for these two yesterday, they were indeed *not* in testing state :) My scripts to generate these lists take the "updates-testing" repo into account, and I manually filtered out any false positives.
Fabio