OSS-Singularity/zcode-dispatch-observatory
OSS-Singularity/zcode-dispatch-observatory
Watch
1
0
Fork
You've already forked zcode-dispatch-observatory
 0
Code Issues Pull requests Projects Releases 137 Packages Wiki Activity Actions

Pinky X: service robustness — reboot-proof steer spawn + baked boot-start (v1.28.0) #54

Merged
ClaudiuSchuster merged 1 commit from oss-oo/pinky-x-robustness into main 2026-09-26 10:14:43 +00:00
Conversation 0 Commits 1 Files changed 8 +714 -45
ClaudiuSchuster commented 2026-09-26 10:07:10 +00:00
Owner

Two defects in one service-robustness round (owner-reported after the 2026-09-26 station reboot):

S-1 — steering broken after reboot. After a reboot the user manager starts zcode-observatory.service with the default PATH, which lacks ~/.local/bin (where the zcode-dispatch launcher symlink lives) — every steer died on the bare shutil.which lookup. Fixed at BOTH layers: the unit ships Environment="PATH=%h/.local/bin:…" (%h home specifier, no hardcoded username), and the steer spawn resolves the launcher through an explicit candidate order (resolve_dispatcher): ZDO_DISPATCH_BIN override (missing override = loud error, never a silent fall-through) → absolute ~/.local/bin/zcode-dispatch → the tracked install dirs (sibling checkout, ~/tools/zcode-dispatch, ~/.local/share/zcode-dispatch) → bare PATH lookup last resort. The 503 refusal lists every candidate tried. Both spawn sites use it (start_steer + fastlane delivery).

S-2 — boot-start baked into install/update (owner decision 2026-09-26): install.sh runs systemctl --user enable --now after daemon-reload; the unit's outdated "deliberately NOT auto-enabled" comment is gone; README documents boot-start and the resolution order.

Tests (84, +15): resolution-precedence units; a real boot-env integration rig (server under the scrubbed fresh-boot environment, free port — steer spawns via the absolute candidate, override wins); station-unit contract updated (Environment line + enablement); v1.27.0 validation follow-up F1 — the missing invalidation mutation tests (stream-history delivered flag, stream-dropped rewrite, provider-config touch, live-driver/alive-steer digest flips; each asserts the next poll reflects it).

Local gates: full suite 84/84 green; Super-Linter station recipe green (BASH, BASH_EXEC, MARKDOWN, PYTHON_PYLINT, PYTHON_RUFF, YAML); G5 CPU envelope re-check on a 120-session fixture rig: 0.72 % of one core over 22 poll cycles (pre-v1.27.0 failure mode ~170 %); fixture CDP probes green (page builds, version marker v1.28.0).

Two defects in one service-robustness round (owner-reported after the 2026-09-26 station reboot): **S-1 — steering broken after reboot.** After a reboot the user manager starts `zcode-observatory.service` with the default PATH, which lacks `~/.local/bin` (where the `zcode-dispatch` launcher symlink lives) — every steer died on the bare `shutil.which` lookup. Fixed at BOTH layers: the unit ships `Environment="PATH=%h/.local/bin:…"` (%h home specifier, no hardcoded username), and the steer spawn resolves the launcher through an explicit candidate order (`resolve_dispatcher`): `ZDO_DISPATCH_BIN` override (missing override = loud error, never a silent fall-through) → absolute `~/.local/bin/zcode-dispatch` → the tracked install dirs (sibling checkout, `~/tools/zcode-dispatch`, `~/.local/share/zcode-dispatch`) → bare PATH lookup last resort. The 503 refusal lists every candidate tried. Both spawn sites use it (`start_steer` + fastlane delivery). **S-2 — boot-start baked into install/update** (owner decision 2026-09-26): `install.sh` runs `systemctl --user enable --now` after `daemon-reload`; the unit's outdated "deliberately NOT auto-enabled" comment is gone; README documents boot-start and the resolution order. **Tests (84, +15):** resolution-precedence units; a real boot-env integration rig (server under the scrubbed fresh-boot environment, free port — steer spawns via the absolute candidate, override wins); station-unit contract updated (Environment line + enablement); v1.27.0 validation follow-up **F1** — the missing invalidation mutation tests (stream-history delivered flag, stream-dropped rewrite, provider-config touch, live-driver/alive-steer digest flips; each asserts the next poll reflects it). Local gates: full suite 84/84 green; Super-Linter station recipe green (BASH, BASH_EXEC, MARKDOWN, PYTHON_PYLINT, PYTHON_RUFF, YAML); G5 CPU envelope re-check on a 120-session fixture rig: 0.72 % of one core over 22 poll cycles (pre-v1.27.0 failure mode ~170 %); fixture CDP probes green (page builds, version marker v1.28.0).
Pinky X: service robustness — reboot-proof steer spawn + baked boot-start (v1.28.0)
All checks were successful
Checks / Functional checks (pull_request) Successful in 18s
Details Explain this check
1e917568f6 
S-1 (steering broken after reboot, owner-hit 2026-09-26): after a reboot
the user manager starts the unit with the default PATH, which lacks
~/.local/bin (where the zcode-dispatch launcher symlink lives) — every
steer died on the bare PATH lookup. Belt and suspenders:
- zcode-observatory.service ships Environment="PATH=%h/.local/bin:…"
  (%h home-dir specifier, no hardcoded username)
- observatory.py resolves the launcher through an explicit candidate
  order (resolve_dispatcher): ZDO_DISPATCH_BIN override (a missing
  override is a loud error, never a silent fall-through) >
  ~/.local/bin/zcode-dispatch (Path.home()-based) > the tracked install
  dirs (observatory sibling checkout, ~/tools/zcode-dispatch,
  ~/.local/share/zcode-dispatch) > bare PATH lookup (last resort); the
  503 refusal lists every candidate tried. Both spawn sites use it
  (start_steer + fastlane delivery, which logs the reason on requeue)

S-2 (boot-start baked in, owner decision 2026-09-26): install.sh runs
systemctl --user enable --now after daemon-reload; the unit's outdated
'deliberately NOT auto-enabled' comment is gone; README documents the
boot-start behavior and the resolution order.

Tests (84 total, +15): dispatcher-resolution precedence units; a real
boot-env integration rig (fixture DB + fake HOME, server under the
scrubbed fresh-boot environment on a free port — steer spawns via the
absolute candidate, override wins); station-unit contract updated for
the Environment line + enablement; validation follow-up F1 — the
missing invalidation mutation tests (stream-history delivered flag,
stream-dropped rewrite, provider-config touch, live_driver_digests and
alive_steer_counts flips; each asserts the next poll reflects it).
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
OSS-Singularity/zcode-dispatch-observatory!54
No description provided.