[huong8373tt-beep]

Summary

• let report writers return an empty files list when a retry finds content already committed locally
• route those retries through the existing unpushed-commit push path instead of reporting already-present
• cover the shared report path and a real session save --push retry after a rejected push

Fixes #1277

Validation

• npx vitest run src/__tests__/git-kind-reports.test.ts -t "pushes an earlier unpushed report commit when the retry has no new files"
• npm run build; npx vitest run --config vitest.e2e.config.ts src/__tests__/e2e/reports-writer-sync-561.test.ts -t "pushes an earlier rejected session summary when retrying an existing session"
• npm run typecheck
• npm run lint
• git diff --check

Slai Agent Runtime assisted with investigation, implementation, and testing. The account owner reviewed and authorized this contribution.

[huong8373tt-beep]

Closing this duplicate in favor of #1280.

#1280 already includes the #1277 report-retry fix in its branch: it changes the shared reports writer so an empty files result still enters the existing unpushed-commit push path, and its real CLI evidence covers a rejected session save --push followed by a retry. I verified that the #1280 diff touches src/save-session.ts, src/utils/branch-worktree.ts, and the shared report regression test; keeping #1291 would create overlapping work.

Thank you for the issue and implementation context.

[huong8373tt-beep]

Correction to my close note: #1280 only partially overlaps #1277, so this is not a duplicate.

#1280's own description marks #1277 as related but outside its scope. Its diff changes the shared push mechanics and session save --push, but it does not update the remaining #1277 writers that return null when their content is already committed locally: bootstrap/member registration, both init paths, vote sync, and automatic reports. This PR covers those remaining paths as well as the shared retry behavior.

Reopening #1291 to keep the full #1277 fix available. Sorry for the incorrect duplicate assessment.

[github-actions]

No findings in the specified PR diff.

The description documents sufficient testing, including a build followed by a representative real-CLI E2E test for retrying a rejected session-summary push. I did not run, build, or install anything from the PR.

[SaulMoro]

This was generated by AI during triage.

@huong8373tt-beep Thanks, this does what #1277 asks. We built b87d8a9e and origin/main (0d57ac46) and checked it against the issue.

#1277	Result
A retry whose content is already committed pushes the earlier commit	Reproduced with session save --push against a local origin that rejects the first push. On main the retry prints Session already present in the team log — nothing to push. and origin never gets the commit. On this PR it prints Pushed a previously unpushed session summary. and origin has sessions/alice/2026-10.md.
bootstrap.ts, both init.ts sites, both votes sites in hook-handlers.ts, and the team-push.ts auto-report return { files: [] }	Checked by reading the code. A local fixture can't re-run init or bootstrap into the retry.
Read-only callbacks keep null and never push	Every updateReports caller checked; reports-branch-readonly.test.ts passes

We also checked that nothing local is dropped when the retry fails again. Usage events, votes deltas and the member record are cleared only after origin confirms the push. A pull whose reports push was rejected kept its 3 usage events and cleared them only after the next run delivered.

tsc, lint and git diff --check pass, the related unit tests pass, and your e2e passes on this head.

Non-blocking

[P2 non-blocking] src/save-session.ts:163-164: when the retry is rejected again, session save --push still prints Session already present in the team log — nothing to push. and exits 0. isPublished is false for failed and busy as well as for already-present. Main prints the same line, but there it was true, because no push was attempted. A later teamai pull does deliver the commit. A small fix would be:

const result = await withTimeout(updateReports(/* unchanged */), 10_000, 'Push timeout (10s)');
const pushed = isPublished(result);
// in the `ran && !rel` branch:
if (pushed) spin.succeed('Pushed a previously unpushed session summary.');
else if (result.status === 'failed') spin.warn('Push failed. The summary is saved locally; retry with: teamai session save --push');
else spin.info('Session already present in the team log — nothing to push.');

[P2 non-blocking] No unit test asserts the new { files: [] } returns in bootstrap, init, votes sync and judge, or the auto-report. Only the shared path and session save are covered. A test for one of them, ideally the votes sync, would keep these from going back to null.

[P3 nit] The body lists the test commands but not what passed. Adding the summary lines would make it a record.

Note: #1280 changes the same updateImpl branch in src/utils/branch-worktree.ts, where the two versions behave the same, and the same saveSession branch, where it uses different messages. Whichever lands second has to reconcile save-session.ts and the assertion at src/__tests__/e2e/reports-writer-sync-561.test.ts:237.

[github-actions]

No findings in the specified PR diff.

The PR description documents sufficient testing: a build followed by a representative real-CLI E2E test covering a session-summary retry after a rejected push. No additional provider × agent coverage is required.

The earlier review reported no findings, so there are none to mark resolved. I did not run, build, or install anything from the PR.

[huong8373tt-beep]

Thanks for the approval and the specific follow-up. I addressed the first non-blocking item in 750e29a:

• session save --push now distinguishes an already-present retry from a failed or busy retry after the summary is already committed locally.
• A failed retry reports that the summary remains local and gives the retry command; a busy retry does the same with the branch-busy reason.
• The existing real-CLI rejected-push/retry test now covers the failed-again message before the next retry delivers the pending commit.

Validated on this head: npm run build, the targeted built-CLI Vitest case, npm run typecheck, npm run lint, and git diff --check.

I am leaving the suggested per-writer fixtures as follow-up rather than add speculative fixtures in this focused patch; the shared writer behavior and the reviewed call-site conversions remain covered by the existing regression.