From 27a1c39c59cc46d8d996fb3ef3a5ac8ae785511d Mon Sep 17 00:00:00 2001 From: kami Date: Fri, 28 Aug 2026 14:49:39 +0400 Subject: [PATCH] orchestra: TASK.md --- TASK.md | 63 +++++++++++++++++++++++++-------------------------------- 1 file changed, 27 insertions(+), 36 deletions(-) diff --git a/TASK.md b/TASK.md index 1ce3b8e..7de7613 100644 --- a/TASK.md +++ b/TASK.md @@ -1,51 +1,42 @@ -# Task 06G4FT0MB733H2ZCZ237WFHFD8 +# Task 06G4FZA825XMTCQ8ME8NN82YM8 -Give the healthcheck a self-test and a machine-readable mode +Add a --strict mode to the healthcheck and keep the pinned usage text honest - Project: test-e2e -- Source: gitea:test-e2e/10 +- Source: gitea:test-e2e/12 - Priority: 0 ## Instructions -Three pieces of work, in order. Plan them as exactly three phases in that order. +Three phases, in this order. -## Phase 1 — a self-test for the healthcheck +## Phase 1 — `--strict` on the healthcheck -New file `scripts/orchestra_e2e_selftest.sh`, executable, `#!/usr/bin/env bash`. +`scripts/orchestra_e2e_healthcheck.sh --strict` must reject an unrecognised flag: print `unknown flag: ` on stderr and exit 2. Without `--strict` the current silent-ignore behaviour is unchanged, and `--help` still wins over everything. -It runs `scripts/orchestra_e2e_healthcheck.sh`, then asserts: +Automated verification only: -- exit status is 0 -- stdout contains `OK` - -On failure it prints which assertion failed and exits 1. On success it prints `selftest: ok` and exits 0. - -Verification for this phase is automated only. The one check is: - - - run: ["bash", "-n", "scripts/orchestra_e2e_selftest.sh"] - -Do not add a manual step to this phase. - -## Phase 2 — `--json` on the healthcheck - -`scripts/orchestra_e2e_healthcheck.sh --json` prints exactly one line: - - {"status":"ok","checks":1} - -and exits 0. Without `--json` the existing human output is unchanged, and `--help` keeps working. `--json` and `--help` together behave as `--help`. - -Verification for this phase must include both an automated check and a manual step. The JSON shape is a judgement a human confirms by reading it, so state the manual step as the exact command a human runs and the exact line they should see. - -## Phase 3 — teach the self-test about `--json` - -Extend `scripts/orchestra_e2e_selftest.sh` to run the healthcheck twice: once plain and once with `--json`. Assert the plain run still contains `OK`, and assert the `--json` run prints one line beginning with `{"status":`. - -Verification for this phase is automated only: - - - run: ["bash", "-n", "scripts/orchestra_e2e_selftest.sh"] + - run: ["bash", "-n", "scripts/orchestra_e2e_healthcheck.sh"] - run: ["bash", "scripts/orchestra_e2e_healthcheck.sh"] +## Phase 2 — cover `--strict` in the self-test + +Extend `scripts/orchestra_e2e_selftest.sh` so it asserts that `--strict --nonsense` exits 2 and that `--strict` alone still exits 0. + +Automated verification only: + + - run: ["bash", "-n", "scripts/orchestra_e2e_selftest.sh"] + +## Phase 3 — keep the pinned usage text honest + +`scripts/test_healthcheck.sh` pins the healthcheck's exact `--help` output in a hardcoded `USAGE` variable and compares with string equality. Phase 1 adds `--strict`, so that pinned text must be updated in place to match, and the pinned comparison must keep passing. + +Work from the `USAGE` variable that is already in `scripts/test_healthcheck.sh`. Do not rewrite the file's structure. + +Automated verification for this phase must run the pinned test itself: + + - run: ["bash", "scripts/test_healthcheck.sh"] + ## Constraints The project's verification policy allows exactly two command shapes: @@ -57,7 +48,7 @@ A `run:` line outside those is refused when the plan seals, not later. ## Research first -Before planning, establish at least three findings and label their confidence honestly: at least one `fact`, one `inference`, and one `assumption`. The plan must cite at least two of them in References. +Establish at least three findings with distinct confidence: at least one `fact`, one `inference`, one `assumption`. Cite at least two in References. ## Completion