Skip to content

fix(gdm): accept gnome-keyring-pam until oo7 publishes - #1194

Merged
Danathar merged 2 commits into
projectbluefin:mainfrom
juanresendiz813:fix/gdm-pam-oo7-fallback
Nov 18, 2026
Merged

Danathar merged 2 commits into
projectbluefin:mainfrom
juanresendiz813:fix/gdm-pam-oo7-fallback

Conversation

@juanresendiz813

Copy link
Copy Markdown
Contributor

Summary

Every full factory run is currently refused at publish, and the cause is gdm, not the mirror timeouts in the same run.

#1156 changed gdm to Requires: pam_oo7, but oo7 has never built here, so nothing provides pam_oo7. The scheduled run 44010051148 rebuilt gdm 1:51.0-2, and the publish job's Hummingbird-only transaction then refused the whole candidate:

Problem 1: package gnome-shell-51.1-1.hum1.bfin.x86_64 from factory requires gdm-libs(x86-64), but none of the providers can be installed
  - nothing provides pam_oo7 needed by gdm-1:51.0-2.hum1.bfin.x86_64 from factory
Hummingbird-only consumer transaction is not resolvable

That run built gupnp-av, kmscon, sudo-rs, uutils-coreutils, nerd-fonts-symbols, google-noto-serif-cjk-vf-fonts and gnome-shell-extension-blur-my-shell. None of them published, because the gate failed. The intermediate gate (publish0) failed on the same pam_oo7 line. #1192's triage comment blames the mirror outage for the whole run. The outage did cause the baobab, rygel and glog failures, but re-running cannot fix the gate: every full run that rebuilds gdm will fail it the same way until pam_oo7 exists.

oo7 itself fails at lock resolution: nothing provides requested (crate(formatx/default) >= 0.4.0 ...), and the same for gettext-rs 0.8 and zgvariant 1. Fedora 44 has formatx 0.2.4 and gettext-rs 0.7.7, and no zgvariant at all. A separate PR vendors those crates.

Change

  • packages/gdm/gdm.spec: Requires: (pam_oo7 or gnome-keyring-pam). gdm 51's data/pam-redhat/gdm-password.pam and gdm-autologin.pam load both pam_gnome_keyring.so and pam_oo7.so as -auth/-password/-session optional, so either provider works. The published gnome-keyring-pam 50.0-1 satisfies it today, and pam_oo7 satisfies it once oo7 builds. Release stays 2, because 1:51.0-2 never published (the published gdm is 1:51.0-1).
  • .agents/skills/build-failure-triage/SKILL.md: a classification row for this signature (a Requires on a factory subpackage that has never published blocks the whole candidate).

One side note for whoever lands utah's oo7 switch: oo7's Obsoletes: gnome-keyring(-pam) < 51 sits behind %if ! (0%{?fedora} && 0%{?fedora} < 45). The build root has fedora-release 44, so that Obsoletes is not emitted here, and pam_oo7 and gnome-keyring-pam can be installed together.

Validation

  • just check: green (1650 passed).
  • rpmspec -q --requires in fedora:44 prints (pam_oo7 or gnome-keyring-pam).
  • Not built locally. The gate should pass once a full run rebuilds gdm with this change, but I haven't verified that.

Refs #1192, #1156, projectbluefin/utah#1313

Assisted-by: Claude Opus 5.5 via Claude Code

@hivecommons-hive hivecommons-hive Bot added author/human A person pushed this PR themselves no-dco A commit is missing Signed-off-by — run `git rebase --signoff` role/untrusted Author has no registry profile or verified hive claim (or is revoked); never auto-merged stage/050-review-pending Review swarm requested; awaiting a verdict for the current head labels Nov 17, 2026
projectbluefin#1156 made gdm require pam_oo7, but oo7 has never built here: its
Rawhide snapshot needs formatx 0.4, gettext-rs 0.8 and zgvariant 1, and
Fedora 44 provides none of them. The scheduled run 44010051148 rebuilt
gdm 1:51.0-2, and the publish job's Hummingbird-only transaction then
refused the whole candidate on "nothing provides pam_oo7 needed by gdm".
gupnp-av, kmscon, sudo-rs, uutils-coreutils, nerd-fonts-symbols,
google-noto-serif-cjk-vf-fonts and gnome-shell-extension-blur-my-shell
all built in that run and are still unpublished because of it.

gdm 51's PAM files load both pam_oo7.so and pam_gnome_keyring.so as
optional modules, so either provider works. Require either one, so gdm
installs with the published gnome-keyring-pam now and with pam_oo7 once
oo7 builds. The triage skill gets the failure signature.

Assisted-by: Claude Opus 5.5 via Claude Code
Signed-off-by: Juan Resendiz <juanresendiz813@gmail.com>
@juanresendiz813
juanresendiz813 force-pushed the fix/gdm-pam-oo7-fallback branch from a45262b to 72947bd Compare November 17, 2026 22:00
@hivecommons-hive hivecommons-hive Bot added dco All commits are DCO signed-off stage/stalled In its current stage longer than expected stage/060-reviewed Review swarm verdict approve; a gate is still pending (CI running, soak not started) stage/090-ready-for-human-merge Hive would merge this but trust/role blocks it; a maintainer's /lgtm /approve is all that is missing and removed no-dco A commit is missing Signed-off-by — run `git rebase --signoff` stage/050-review-pending Review swarm requested; awaiting a verdict for the current head stage/stalled In its current stage longer than expected stage/060-reviewed Review swarm verdict approve; a gate is still pending (CI running, soak not started) labels Nov 17, 2026
@hivecommons-hive

Copy link
Copy Markdown
Contributor

Ready for a human merge: verdict approved @72947bd, 23/23 checks green, soak 30m0s elapsed. Blocked only by role/untrusted (no verified hive claim or registry tier).

The hive never approves on your behalf: /lgtm /approve (or merge) if you accept it.

@hivecommons-hive hivecommons-hive Bot added stage/060-reviewed Review swarm verdict approve; a gate is still pending (CI running, soak not started) stage/090-ready-for-human-merge Hive would merge this but trust/role blocks it; a maintainer's /lgtm /approve is all that is missing and removed stage/090-ready-for-human-merge Hive would merge this but trust/role blocks it; a maintainer's /lgtm /approve is all that is missing stage/060-reviewed Review swarm verdict approve; a gate is still pending (CI running, soak not started) labels Nov 18, 2026

@hanthor hanthor left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The change is right, and it unblocks publish: no repository provides pam_oo7 today, so every full run that rebuilds gdm is refused. gdm 51's gdm-password and gdm-autologin load both pam_gnome_keyring.so and pam_oo7.so as optional, so either provider works. Keeping Release 2 is fine because 1:51.0-2 never published.

How this fits with projectbluefin/utah#1304 (oo7 replaces gnome-keyring):

  • With this PR, utah#1304 no longer needs a gdm that requires pam_oo7. It only needs oo7 to publish. Its body says the published gdm conflicts with pam_oo7 through Obsoletes: gnome-keyring-pam < 51, but, as you noted, that Obsoletes is not emitted on the F44 build root. The two can be installed together.
  • So utah#1304 must keep gnome-keyring-pam (and gnome-keyring) in [excluded]. If it doesn't, the image could end up with both PAM modules and two keyring daemons unlocking at login.
  • Follow-up: once oo7 publishes and utah#1304 lands, go back to Requires: pam_oo7 to match Rawhide, or keep the rich dependency on purpose and say so in the spec comment.

This touches a PAM dependency, so I'm leaving the merge to the maintainer and not approving here.

Reviewed on behalf of @hanthor with Claude Code

@Danathar
Danathar merged commit ec4850f into projectbluefin:main Nov 18, 2026
28 checks passed
@hanthor

hanthor commented Nov 18, 2026

Copy link
Copy Markdown
Member

Post-merge security check: this is fine. In gdm 51's gdm-password and gdm-autologin, both pam_gnome_keyring.so and pam_oo7.so are -auth/-password/-session optional. A missing module is skipped, and neither module decides whether a login succeeds; the gdm-password-auth-substack and pam_gdm lines do. So whichever provider the rich dependency pulls in, the login stack fails safe.

One thing to watch: if an image ever ships both, both keyring daemons auto-start at login. Utah should pick one.

Reviewed on behalf of @hanthor with Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

author/human A person pushed this PR themselves dco All commits are DCO signed-off role/untrusted Author has no registry profile or verified hive claim (or is revoked); never auto-merged stage/090-ready-for-human-merge Hive would merge this but trust/role blocks it; a maintainer's /lgtm /approve is all that is missing

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants