[test] Sessions-rework release QA: plan, findings, gate extensions - #5546
[test] Sessions-rework release QA: plan, findings, gate extensions#5546mmabrouk wants to merge 7 commits into
Conversation
…fresh, gate driver extensions - QA plan + execution log for the v0.106 sessions release at docs/design/agent-workflows/projects/qa/release-2026-07-sessions-storage-rework.md - agent-release-gate skill: fix stale fingerprint/agentRequest/STATUS.md references, document the four session flags and the differential QA method - qa_product.py: --last-message-only differential mode with sent_messages dumps, new records/sessions/followup journeys Claude-Session: https://claude.ai/code/session_01VhviVuZExGJUnY28mzbPBK
…tion procedure Encodes the v0.106 sessions release QA method: branch map + harness audit first, plan doc with execution log, layered phases (build/migration gate, flag-state wire runs with differential diffs, mismatch cells, REST + acceptance, depth probes, recorded browser pass), stack-state discipline, and evidence rules. Claude-Session: https://claude.ai/code/session_01VhviVuZExGJUnY28mzbPBK
…disclosure SKILL.md stays lean; the plan-doc template and a sanitized worked example ship inside resources/ and load only at the step that needs them. Drops the external docs-path dependency so the skill is shareable as one folder. Claude-Session: https://claude.ai/code/session_01VhviVuZExGJUnY28mzbPBK
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Important Review skippedDraft detected. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
QA recording (watch first): qa-ui-pass-sessions-rework.mp4 — a 39-frame screenshot walkthrough of the browser pass (session rail, always-allow, warm Stop, Steer, the paused-gate-refresh defect, config drawer, Pi multi-gate approvals). Hosted as a draft-release asset; repo access required. Per-scenario verdicts and evidence are in the plan doc's execution log in this PR. Defects found in the browser pass: #5542 (paused gate lost on refresh), #5543 (young-delete resurrection), #5544 (config drawer what-changed), #5545 (batch-approval UI unreachable). |
1 participant
Context
feat/sessions-storage-reworkbundles the sessions storage rework plus the feature train stacked on it (server session list, concurrent approvals + batch UI, warm Stop and Steer, config drawer rework), about 600 files. Before release we needed to know what actually ships, whether the always-on surface regressed, and whether the flag-gated sessions train holds up beyond unit tests. The existing agent-release-gate skill only covered the flags-off legacy path and had gone stale in three places.Changes
docs/design/agent-workflows/projects/qa/release-2026-07-sessions-storage-rework.md. Every phase carries its verdict and evidence; the verdict summary at the bottom lists the issues found.agent-release-gateskill refresh: the stale fingerprint invariant (now conditional on client mode), the outdated "frontend always echoes full history" claim, and dangling STATUS.md pointers are fixed; a new addendum documents the four session flags, their per-layer truthiness parsing, and the differential QA method.qa_product.py): a--last-message-onlymode that mirrors the frontend's minimal-history behavior and dumps the exact per-turn messages for offline diffing, plus three new journeys:records(durable-record readback with ordering and truncation assertions),sessions(REST lifecycle: query, archive, unarchive, rename, delete), andfollowup(post-approval-resume turn checking for duplicated tool-call records).release-qaskill encoding the whole procedure (branch map, plan template, layered phases, stack-state discipline, evidence rules), with the template and a sanitized worked example as progressive-disclosure resources.QA results (summary)
All wire, acceptance, migration, build, and depth checks passed. Full log in the plan doc. Findings filed as issues: #5542 (refresh while paused on an approval gate loses the gate; the sharpest one), #5543 (young-session delete resurrects), #5544 (config drawer what-changed gaps), #5545 (batch-approval UI unreachable on every harness), #5538 (keepalive teardown race after a same-session takeover), #5539 (concurrent-turn design characterization), #5534 (run.sh env-file tooling bug).
Tests / notes
https://claude.ai/code/session_01VhviVuZExGJUnY28mzbPBK