fix: detached docker session stays executing while container is running

[konard]

Summary

Fixes #136. A detached --isolated docker session was recorded with a terminal status (executed) and exitCode -1 while its container was still running — or not even created yet on a slow Docker-in-Docker host. This made $ --list / $ --status report the session as finished/failed when it was actually still in progress, breaking orchestrators that poll status.

Root cause

isDetachedSessionAlive (JS) / is_detached_session_alive (Rust) ran docker inspect -f '{{.State.Running}}' <name> and treated any failure (non-zero exit / empty output) as "not alive" (false). Right after docker run -d returns, the container can be transiently invisible to docker inspect on a slow DinD host. A false liveness result drove enrichDetachedStatus / enrich_detached_status to flip the record to executed and write the -1 sentinel exit code.

Fix (JS + Rust parity)

• Distinguish "unknown" (inspect failed → null/None) from "stopped" (inspect succeeded, Running=false). When liveness is unknown, the detached session stays executing instead of fabricating a terminal -1 result.
• Inspect now reads both {{.State.Running}} and {{.State.ExitCode}} in one call. When a container has genuinely stopped, the real exit code is resolved from the log footer first, then from docker inspect .State.ExitCode, only falling back to -1 as a last resort when no real code is available.
• A removed/--rm container that already wrote an Exit Code: log footer still resolves to its real terminal status (footer is authoritative when liveness is unknown).

How to reproduce

1. Start a long-running detached docker session: $ --isolated docker --detached -- sleep 30.
2. Immediately run $ --status <session> (or $ --list) while the container is still spinning up / running.
3. Before: status executed, exit code -1. After: status executing until the container actually exits, then the real exit code.

Tests

• JS (js/test/session-name-status.js): added Issue #136: detached docker session liveness — unknown container → null; not-visible container keeps executing; running container stays executing; stopped container resolves the real exit code. Guarded on docker availability.
• Rust (rust/tests/session_name_status.rs): added 4 parity tests, guarded on docker availability.
• All JS and Rust suites pass; eslint, prettier --check, cargo fmt --check, and cargo clippy are clean.

Changelog

• js/.changeset/detached-docker-running-status.md (start-command: patch)
• rust/changelog.d/136.md (bump: patch)

[konard]

Working session summary

The working tree is clean (ci-logs is gitignored or already absent). Everything is complete. Let me give a final summary.

The work on issue #136 is complete and verified.

Summary

Issue #136: A detached --isolated docker session was reported with a terminal status (executed) and exitCode -1 while its container was still running — or not yet created on a slow Docker-in-Docker host — breaking orchestrators that poll $ --status / $ --list.

Root cause: isDetachedSessionAlive (JS) / is_detached_session_alive (Rust) ran docker inspect -f '{{.State.Running}}' and treated any failure as "not alive" (false). Right after docker run -d, the container can be transiently invisible to docker inspect. That false liveness drove the enrich step to flip the record to executed with the -1 sentinel.

Fix (JS + Rust parity):

• Distinguish unknown (inspect failed → null/None) from stopped (inspect succeeded, Running=false). Unknown liveness keeps the session executing instead of fabricating a terminal result.
• Inspect now reads {{.State.Running}} and {{.State.ExitCode}} in one call. A genuinely stopped container resolves its real exit code from the log footer, then docker inspect .State.ExitCode, only falling back to -1 as a last resort.

Tests: 4 new JS tests + 4 new Rust tests, each guarded on docker/Linux-image availability. A follow-up commit fixed a Windows-CI flake by probing with the repo's canRunLinuxDockerImages() and skipping the stopped-container test unless the container actually exists.

Verification:

• JS suite, Rust suite (28 in the relevant file), eslint, prettier --check, cargo fmt --check, cargo clippy — all clean.
• Both CI workflows (JavaScript + Rust) pass on the final commit 2f15524.
• Changelog fragments added: js/.changeset/detached-docker-running-status.md and rust/changelog.d/136.md.

PR #137 is updated (title de-WIP'd, full description), marked ready for review, and MERGEABLE: #137

---

This summary was automatically extracted from the AI working session output.

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

• Public pricing estimate: $5.802376
• Calculated by Anthropic: $6.369713
• Difference: $0.567337 (+9.78%)

📊 Context and tokens usage:

Claude Opus 4.8: (2 sub-sessions)

1. 116.7K / 1M (12%) input tokens, 33.1K / 128K (26%) output tokens
2. 59.7K / 1M (6%) input tokens, 13.0K / 128K (10%) output tokens

Total: (6.2K new + 153.6K cache writes + 7.0M cache reads) input tokens, 51.7K output tokens, $5.802376 cost

🤖 Models used:

• Tool: Anthropic Claude Code
• Requested: opus
• Model: Claude Opus 4.8 (claude-opus-4-8)

📎 Log file uploaded as Gist (2980KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.

[konard]

✅ Ready to merge

This pull request is now ready to be merged:

• All CI checks have passed
• No merge conflicts
• No pending changes

---

Monitored by hive-mind with --auto-restart-until-mergeable flag