Repository navigation
fix(uwelcome): greet only a terminal so login doesn't stall the portals - #1709
hivecommons-hive[bot] wants to merge 2 commits into
Conversation
GDM starts the desktop session through a login shell, which sources /etc/profile.d/uwelcome.sh. uwelcome then asks for the portal before graphical-session.target is up, so xdg-desktop-portal.service fails its dependency and D-Bus holds every portal request for its 120 s activation timeout: file pickers, screenshots and settings portals are dead for about two minutes after login (common#1688, VM serial log of run 43653470297). Only run the greeting when stdout is a terminal. Tests that exercise the greeting guards now run the profile under script(1) for a pty, and a new test pins that a non-terminal login shell does not call uwelcome. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WcWm7tJ5iop2b4GMkKSPAv Signed-off-by: Danathar <Danathar@users.noreply.github.com>
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: Needs approval from an approver in each of these files:Approvers can indicate their approval by writing |
There was a problem hiding this comment.
Looks correct to me. The red E2E check is not caused by this change, and the run is evidence that the fix works.
Code. system_files/shared/etc/profile.d/uwelcome.sh:16 adds [ -t 1 ] ahead of the existing root and UWELCOME_SHOWN guards. A login shell started by GDM has no terminal on stdout, so it skips uwelcome. Interactive terminals and SSH logins have a pty, so they still get the greeting. The existing bats tests now run under script -qec … /dev/null, which gives them a pty. That means each test still checks its own guard instead of passing only because no tty is present. The new test (tests/test_uwelcome_profile.bats, "skipped without a terminal") covers the GDM path.
E2E attribution. These two runs come from the same base at about the same time:
| Run | Has this fix | Failing scenarios |
|---|---|---|
| #1709, run 44372663960 | yes | common_dconf.feature:91 Searchlight, common_desktop_entries.feature:12, common_desktop_entries.feature:24 |
| #1708, run 44372598326 | no | the same 3, plus common_portals.feature:25, common_portals.feature:29, portal_integration.feature:10 |
The portal scenarios fail without this change and pass with it. #1690's body predicted exactly that. The 3 remaining failures (Searchlight dconf, desktop entries, HTML MIME handler) appear on both PRs. They have nothing to do with the uwelcome profile, so they look like a baseline problem on main or in the Dakota test image, not something this PR introduced.
Housekeeping. #1690 and this PR point at the same head (d7fa364), so whichever one lands makes the other a no-op. The no-dco label is still on this PR. A maintainer should check it against the sign-off requirement before merge.
Confidence: 5/5 (safe)
— hive: agent=reviewer-2 backend=claude model=claude-opus-5-5 effort=medium claude=2.1.326
🐝 Possible duplicateThis PR changes exactly the same files as #1690, which was opened first:
The diff here is byte-identical to #1690. Evidence — identical changed-file set, identical diff (3 file(s)):
This is a suggestion, not a verdict. Hive will not close, label, approve or merge anything here — a human decides. It was produced by comparing changed-file sets across open PRs, which is a candidate generator: PRs that share a file are not necessarily the same work. In this case the patches are byte-identical, which is the strongest corroboration available without reading intent — but confirm before closing anything. |
|
Closing as a duplicate of #1690. This branch is #1690's head |
# bluefin-common PR ## What does this change? Ship Compass launcher defaults, a native Super+Space shortcut, a dedicated GNOME Homebrew manifest and conditional hidden-session startup, and remove obsolete Searchlight configuration. - Enable `compass@tunaos.org` in the GNOME schema default and preserve the other extensions. - Move input-source switching to `XF86Keyboard` / `<Shift>XF86Keyboard`. - Register the native `compass toggle` shortcut without replacing `custom0`–`custom4`. - Ship the dedicated `ublue-os/experimental-tap/compass` Brewfile in the GNOME-specific overlay. - Ship `compass.service` with the absolute Homebrew executable, an installation condition, graphical-session lifetime, D-Bus dependency and ordering after package reconciliation, plus its user preset. - Replace text-only Compass assertions with one compiled-schema/dconf behavior check and correct the documented distinction between preset policy and activation. ## Why? Requested Compass integration, kept strictly in common. Related image-side integration is tracked separately in projectbluefin/dakota#1709; this PR does not close that issue. No image-repository, Containerfile or BuildStream changes are included. ## Verification Fresh preflight on `af509e85b2c1f31624bee5f5a042c9286c4eb29f`: - `just check` - `pre-commit run --all-files` - `actionlint .github/workflows/*.yml` - Full `just test`: 225 pytest tests and all registered Bats suites passed. Used an isolated environment for the existing Python test dependencies and `FFMPEG_NO_CONTAINER=1` to use the native VP9 encoder rather than the host wrapper's inaccessible `/var/tmp`. No tests were skipped or changed to work around that wrapper failure; the suite retains its existing platform-dependent skips. Additional exercised checks: compiled GNOME defaults with `GSETTINGS_BACKEND=memory`; compiled dconf shortcut values; failure with Compass removed followed by success when restored; `systemd-analyze --user verify`; offline user-preset application; real Homebrew manifest parsing; and the actual session unit starting a resident v0.28.2 engine and successfully handling toggle, hide and shutdown with an isolated socket/config. ## Integration limits Common's payload here is defaults, provisioning and startup integration; the Shell extension packaging is tracked in Dakota projectbluefin#1709. Existing user extension and shortcut selections are not forcibly migrated. Consumers must actually apply the user preset or provide equivalent enablement. Default `start --hidden` still spawns an engine without forwarding `serve --no-hotkey`, so this PR does **not** claim portal-dialog-free startup. The runtime smoke used an isolated configuration with an empty launcher hotkey. A separate `serve --no-hotkey` engine unit followed by `start --hidden` may avoid the request without upstream changes, but that two-unit approach has not been tested here. No user configuration is overwritten. ## Checklist - [x] PR title follows Conventional Commits - [x] `just check` passes - [x] `pre-commit run --all-files` passes - [x] Skill documentation and generated skill catalog updated - [x] Existing documentation links and changed paths checked - [ ] CI is green after push — pending PR checks ## AI attribution Both trailers are present on the commit: ```text Assisted-by: github-copilot/gpt-6.1-sol via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> ``` Signed-off-by: Danathar <berths-purger-0g@icloud.com> Co-authored-by: castrojo <1264109+castrojo@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Danathar <berths-purger-0g@icloud.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: James Reilly <jreilly1821@gmail.com>
Supersedes #1690 (adopted after CI went red and the contributor was unavailable; original commits by @Danathar preserved).
This branch is #1690's head
d7fa364441e8as pushed by @Danathar (forkprojectbluefin/common), unchanged. Last lane reason on the original:optional-check-failing:E2E — composed common suite / e2e / GNOME — common.Hive fix agent: rebase this branch on
main, fix the failing checks / review asks, and push to THIS branch. Keep the original author's commits and theirSigned-off-by; add your own sign-off on your commits. Do not open another PR.