smoke: add AT-SPI and crash-regression coverage for new GNOME extensions

[mrbobbytables]

Summary

Closes #793. Bluefin/Dakota is enabling seven new curated GNOME Shell extensions by default (projectbluefin/common#1087). This adds AT-SPI smoke and crash-regression coverage for them, complementing the presence-only checks already in tests/smoke/features/bluefin_extensions.feature.

Note on base branch: the issue asked for this to be based on/opened against v4, but that branch does not exist in projectbluefin/testsuite (only main). This PR is based on and opened against main instead.

What's added

• tests/smoke/features/bluefin_new_extensions.feature — the exact scenario set from the issue's acceptance criteria, covering:
  • copyous (highest priority): AT-SPI clipboard history popover interaction, rapid clipboard-change stress with plain/multiline/Unicode/large payloads, and post-stress GNOME Shell accessibility + bounded-memory assertions
  • syncthing-toggle: Quick Settings toggle/pill/header label ("Sync Folder"), service start/stop, start-stop-only gsettings check
  • Bluetooth-Battery-Meter: panel icon presence, symbolic level-indicator-color setting, clean behavior with no paired devices
  • quicksettings-audio-devices-hider/renamer: Quick Settings audio menu opens cleanly for each
  • tiling-assistant: window-snapping keyboard shortcuts, gesture soft-pass (gestures can't be synthesized in headless CI), overall Shell responsiveness
  • tailscale-gnome-qs: Quick Settings item presence, daemon running/stopped/unavailable state handling
• tests/smoke/features/steps/bluefin_new_extensions_steps.py — step definitions built on the existing SSH/Shell.Eval/journalctl helper conventions from steps.py and gnome_extensions_steps.py.
• tests/unit/test_bluefin_new_extensions_steps.py — unit coverage for the new pure/host-command-construction helpers (37 tests).

Design notes

Several of these extensions are third-party/community-maintained, so their gsettings schema ids and exact AT-SPI role names can't be reliably guessed ahead of running against a real image. Steps that depend on such unverifiable internals degrade to a printed WARNING + soft pass (matching the existing pattern in gnome_extensions_steps.py for headless AT-SPI limitations), while the core crash-regression assertions always run and fail hard:

• GNOME Shell stays accessible via AT-SPI
• No new gnome-shell journal errors at err..emerg since the scenario started
• No gnome-shell coredump
• Clipboard stress keeps gnome-shell RSS within a generous bound (no runaway/leak-shaped growth)

This preserves the crash-regression value described in the issue (especially for copyous, which is called out as having crashed GNOME Shell in practice) without hard-failing on implementation details we can't verify from this environment.

Testing

• ruff check tests/ --select E,F,W --ignore E501 — clean
• python3 -m pytest tests/unit/ -q — 1564 passed (including the 37 new tests)
• PYTHONPATH=. python3 -m behave --dry-run --no-summary tests/smoke/features/ — new feature's steps all resolve (no undefined/ambiguous steps introduced; confirmed no new undefined-step count vs. base branch)
• python3 scripts/update_coverage_snapshot.py — snapshot doc regenerated for the new feature/suite mapping

— hive: backend=copilot model=claude-sonnet-5

🐝 Hive Agent: contributor | SHA: 4f7c0616

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 72.76190% with 143 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...oke/features/steps/bluefin_new_extensions_steps.py	52.17%	143 Missing ⚠️

📢 Thoughts on this report? Let us know!