Repository navigation
feat(auth): oo7 replaces gnome-keyring, as on Dakota - #1304
Conversation
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: hanthor The full list of commands accepted by this bot can be found here. Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
@hanthor this PR is back in your court ( Review asks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Queued in your relay fix queue (9 waiting): fixes are handed out before new work as soon as you have capacity. |
Danathar
left a comment
There was a problem hiding this comment.
Stand-in review (the hive's reviewers are out of quota). Approving; don't merge until utah-packages#1147 publishes oo7, and land the pin bump with this PR. The red contract is only check-repos on packages not in the stack's old pin 523507b8; rebase onto #1248/#1309. Checked against oo7's source at the factory's pinned commit and Dakota's oo7-passwd-pam.bst. The passwd hook line is identical to Dakota's, and pam_oo7 implements chauthtok. oo7-daemon.service has WantedBy=default.target, so the global enable is valid. oo7-daemon finds ~/.local/share/keyrings/*.keyring and migrates it on first unlock with the login password, leaving a .migrated stamp, so existing login keyrings carry over (one whose password differs from the login password stays locked until prompted). The gnome-keyring removal fails the build rather than cascading. Questions:
- SSH agent: the factory's gnome-keyring 50.0 is built with its ssh-agent, and
gcr-ssh-agent.socketships in gcr on Utah and Dakota, but nothing here enables it or setsSSH_AUTH_SOCK. Which agent do users get afterwards? A one-line check or note would settle it. - Ordering: the factory gdm (#1156, merged) already Requires
pam_oo7, so the pin bump that carries it must also carry oo7, or Utah's resolve breaks.
Not boot-tested.
6c6d549 to
63d981e
Compare
901ae17 to
57c0ee7
Compare
|
Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor. ℹ️ Sentinel: informational, author is trusted (owner), merge is not blocked. Head
No stage 095-pre-merge-soak → 072-returned-to-contributor: conflicting |
There was a problem hiding this comment.
Summary: This change swaps the desktop's password-storage service (the "keyring" that keeps saved Wi-Fi and app passwords) from the old GNOME one to the newer oo7, matching the sister project Dakota. Since the last review, the only new commit pulls in the latest main branch. That merge quietly undid four rows of the generated parity report, so the report now says Utah still runs the old keyring, which contradicts this very change.
What I need from you: Regenerate the parity report (one command, below) and push.
docs-currency
The Merge main into feat/dakota-oo7 commit (4dbbaa4) resolved baselines/GAP.md by taking main's side. baselines/triage.toml kept this branch's side. GAP.md is generated from triage.toml ("Generated by scripts/image-baseline.py gap; do not edit", baselines/GAP.md:3), so the two now disagree at head:
| item | triage.toml at head |
GAP.md at head |
|---|---|---|
| oo7-daemon | :26 "Installed from the factory in place of gnome-keyring…" |
:218 "Dakota's Secret Service; Utah runs gnome-keyring" |
| tailscale | :50 issue #1265, "…ships with dakota-config-tailscale-setup…" |
:260 issue #1236, old reason |
| /usr/bin/ublue-firstboot-date | :134 planned, #1265 |
:57 untriaged |
| ublue-wallpaper-month.service | :135 planned, #1266 |
:65 untriaged |
This also skews the status counts. GAP.md:13 says planned 48 and :15 says untriaged 196. The triage file implies planned 50 and untriaged 194.
just check cannot catch this. check() in scripts/image-baseline.py:416-430 validates triage.toml against the surface data and never compares it with GAP.md.
Fix: run the generator and commit the result. Don't hand-edit the file.
python3 scripts/image-baseline.py gap && git add baselines/GAP.mdOther notes
- The other 16 files in the delta match main exactly (
compare d67dd762...4dbbaa43shows no diff for them). The workflow changes inpost-testing-e2e.ymland the fourcreate-github-app-tokenbumps come from main's own PRs (#1327, #1339, #1337). This PR adds nothing new in those files. They show up in this PR's diff only because the base branchfeat/dakota-fonts(#1303) has not merged main yet. Once #1303 catches up, they drop out. - The
contractjob fails at this head (job 135058737211, head 4dbbaa4). The cause ischeck-repos:No match for argument: oo7-daemon/pam_oo7/oo7-portal/oo7-cli, plus the fonts from #1303. That is the factory-publish blocker the PR body already names (utah-packages#1147, #1156). It is not a regression from the delta. Unit tests passed (1329, OK). - The previous review at 91dff02 recorded no findings, so there is nothing to mark fixed or still present.
No findings from: correctness, security, intent-alignment, style.
Findings not on a changed line
baselines/GAP.md:218: The merge of main reverted this row.baselines/triage.toml:26says oo7-daemon is "Installed from the factory in place of gnome-keyring; gap closes on the next baseline", but the generated report still says "Utah runs gnome-keyring". Rows 57, 65 and 260 and the counts at 13/15 have drifted the same way. Fix: runpython3 scripts/image-baseline.py gapand commit the regenerated file.
Confidence: 3/5 (needs attention) — 1 perspective requested changes
— hive: agent=reviewer-2 backend=claude model=claude-opus-5-5 effort=medium claude=2.1.326
ebe87ba to
ec7187f
Compare
4dbbaa4 to
3c927f0
Compare
|
Summary: This change could not be merged because it now conflicts with the fonts change (#1295) that landed a minute earlier. It needs a rebase onto main, then it can merge again. What I need from you: Rebase onto main and resolve the conflict. hive-merge returned 405 'Pull Request has merge conflicts' right after #1295 merged. This is a human-authored branch, so I did not push to it. 🐝 Hive Agent: -- hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.327 |
There was a problem hiding this comment.
Summary: This change swaps the desktop's password-and-secrets store (the service that keeps saved Wi-Fi passwords, app logins and similar secrets) from the old GNOME one to the newer one Dakota uses. The feature itself still looks right after the latest update from the main branch. But that update broke one of the automated checks this change adds, so the test run on this version fails. The fix is a few lines in one test.
What I need from you: Apply the test fix in the inline comment so CI passes again.
correctness
- The new password-hook test fails on this head (CI
contract→Validate manifests, job 136092075630, headcc994b0). The merge of main placedtest_passwd_stack_switches_keyring_hook_to_pam_oo7_onceinsideStatsRefreshCacheTests(tests/test_desktop_contract.py:1076). At the previous head, the stack had rewritten that class withSERVICES,block()andrun_block()helpers. Main's squash (#1309) kept the original class without those helpers, so the test raisesAttributeError: 'StatsRefreshCacheTests' object has no attribute 'block'attests/test_desktop_contract.py:1170. It is the only error in the run (Ran 1412 tests … FAILED (errors=1)). The fix makes the test self-contained (patch inline). I ran the patched logic againstscripts/configure-services.sh:295-306at this head: it returns 0 twice, rewrites only the keyring line, and fails closed withre-fit the pam_oo7 password hook.
Delta since the last review (3c927f0, approved): The new commit is a merge of main (#1309, #1362, #1363, #1365, #1366). The oo7 parts of the PR came through unchanged: packages, configure-services.sh:271-306, contract, preset and oo7.md. The PR's own baselines/GAP.md and triage.toml edits were dropped because main's re-measure (#1363) already marks oo7 as installed by #1304 (baselines/triage.toml:35-37, baselines/GAP.md:212-214). README and package-contract counts (58 / 312) add up to the 370 contract packages that CI validated. No earlier findings to re-check.
No findings from: security, intent-alignment, style, docs-currency.
Confidence: 3/5 (needs attention) — 1 high finding
-- hive: agent=reviewer-2 backend=claude model=claude-opus-5-5 effort=medium claude=2.1.327
| def test_passwd_stack_switches_keyring_hook_to_pam_oo7_once(self): | ||
| # Hummingbird's shadow-utils /etc/pam.d/passwd; Dakota's | ||
| # bluefin/oo7-passwd-pam.bst adds the same hook after the substack (#1227). | ||
| block = self.block("PASSWD_PAM=") |
There was a problem hiding this comment.
Test fails at this head: self.block / self.run_block do not exist in StatsRefreshCacheTests. The merge with main put this test in StatsRefreshCacheTests, which (since #1309's squash) has no block()/run_block() helpers. CI contract fails with AttributeError: 'StatsRefreshCacheTests' object has no attribute 'block'. Make the test self-contained. I checked this logic against scripts/configure-services.sh:295-306 at this head: idempotent rewrite, and it fails closed on an unexpected stack.
def test_passwd_stack_switches_keyring_hook_to_pam_oo7_once(self):
# Hummingbird's shadow-utils /etc/pam.d/passwd; Dakota's
# bluefin/oo7-passwd-pam.bst adds the same hook after the substack (#1227).
- block = self.block("PASSWD_PAM=")
+ import re
+ import subprocess
+
+ services = (ROOT / "scripts/configure-services.sh").read_text()
+ block = re.search(r"^PASSWD_PAM=.*?^fi$", services, re.S | re.M).group(0)
+
+ def run_block(script):
+ return subprocess.run(
+ ["bash", "-euo", "pipefail", "-c", script], capture_output=True, text=True
+ )
+
hummingbird = (
@@
for _ in range(2):
- result = self.run_block(block, "")
+ result = run_block(block)
self.assertEqual(result.returncode, 0, result.stderr)
@@
passwd.write_text("password include system-auth\n")
- result = self.run_block(block, "")
+ result = run_block(block)
self.assertNotEqual(result.returncode, 0)Install the factory's oo7-daemon, pam_oo7, oo7-portal and oo7-cli and remove gnome-keyring(-pam). gdm 51's PAM stacks and gnome-session 51's gnome-portals.conf already name pam_oo7 and oo7-portal, so no edit is needed there. Point shadow-utils' passwd stack at pam_oo7 so password changes re-encrypt the login keyring, as Dakota's oo7-passwd-pam.bst does. oo7-daemon starts with the session, as GNOME OS wires it. Closes #1227 Assisted-by: Claude Opus via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: James Reilly <jreilly1821@gmail.com>
cc994b0 to
e41efb0
Compare
|
Pre-merge: rebased onto main after #1297 (one commit, |
Factory:
oo7and the gdm built againstpam_oo7(projectbluefin/utah-packages#1147, #1156) publish at the c288398 pin (#1347);check-repo-availability.pyresolves the full set withgnome-keyring/gnome-keyring-pamexcluded. Rebased onto main after #1295 (one commit).What changed
[parity]and[factory].parityaddoo7-daemon,pam_oo7,oo7-portalandoo7-cli(gnome-build-metacore-deps/oo7-{daemon,pam,portal}.bst).packages/base.toml[excluded]addsgnome-keyringandgnome-keyring-pam.gdm-passwordandgdm-autologin(data/pam-redhat) already carry-auth/-password/-session optional pam_oo7.so, the same lines as GNOME OS'sfiles/gdm/*.pam. The contract assertspam_oo7.soin/usr/lib/pam.d/gdm-password.bluefin/oo7-passwd-pam.bstmakespasswdand Settings re-encrypt the oo7 login keyring. Hummingbird's shadow-utils/etc/pam.d/passwd(checked inos-minimal) already haspassword substack system-authfollowed by-password optional pam_gnome_keyring.so use_authtok.configure-services.shpoints that line atpam_oo7.so. The edit is guarded: already switched is a no-op, and any other shape fails the build.gnome-portals.confalready setsorg.freedesktop.impl.portal.Secret=oo7-portal;gnome-keyring;, which the contract checks. No Utah config is needed.oo7-daemon.serviceintognome-session.target.wants. Utah enables it globally: user preset,systemctl --global enableinconfigure-services.sh(required, so the build fails if the unit is missing), and the contractuser_enabled.desktop-contract/references/oo7.md, a link fromdesktop-contract.md, and triage updates. A new test covers the passwd hook (idempotent, fails closed). The Flathub-descriptor fixture gains the oo7 user unit.Closes #1227
Verification
umask 022; just checkpassed.pre-commit runon the changed files passed.