Files
Agent-System-Dev/tests/test_task_recovery.py
T

246 lines
12 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
"""Local recovery is task-scoped and cannot grant delivery readiness."""
import json
import os
import sys
import unittest
# 允许 `python3 -m unittest tests/test_task_recovery.py` 直接运行(与 discover 模式一致地导入同目录夹具)。
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
import test_task as lifecycle # noqa: E402
class TaskRecoveryTests(unittest.TestCase):
def setUp(self):
self.fixture = lifecycle.TaskLifecycleTests(methodName="runTest")
self.fixture.setUp()
self.addCleanup(self.fixture.doCleanups)
self.wt = self.fixture.worktree()
self.card = self.fixture.record_task_card(self.wt)
self.receipt_number = 0
def precheck(self, mode="local", cwd=None):
f = self.fixture
return f.task(cwd or self.wt, "precheck", check=False,
env=f.env | {"TASK_WORK_MODE": mode})
def capture(self, mode="local"):
f = self.fixture
self.receipt_number += 1
path = f.root / f"recovery-{self.receipt_number}.json"
result = f.run_cmd(self.wt, "bash", "scripts/evidence.sh", check=False,
env=f.env | {"TASK_WORK_MODE": mode, "RECEIPT": str(path),
"CMD": "make precheck"})
self.assertEqual(result.returncode, 0, result.stdout)
return path
def resume(self, receipt, mode="local", cwd=None):
f = self.fixture
return f.task(cwd or self.wt, "resume", check=False,
env=f.env | {"TASK_WORK_MODE": mode, "PRECHECK_RECEIPT": str(receipt)})
def save_card_change(self):
f = self.fixture
f.git(self.wt, "add", str(self.card.relative_to(self.wt)))
f.git(self.wt, "commit", "-m", "Update fixture registration")
def test_local_precheck_allows_known_lag_without_changing_head(self):
f = self.fixture
head = f.git(self.wt, "rev-parse", "HEAD").stdout
f.advance_remote()
strict = self.precheck(mode="delivery")
self.assertNotEqual(strict.returncode, 0)
self.assertIn("SYNC_REQUIRED", strict.stdout)
local = self.precheck()
self.assertEqual(local.returncode, 0, local.stdout)
self.assertIn("[verdict] LOCAL_READY", local.stdout)
self.assertNotIn("[verdict] READY", local.stdout)
self.assertEqual(f.git(self.wt, "rev-parse", "HEAD").stdout, head)
def test_local_precheck_never_advertises_delivery_even_when_current(self):
result = self.precheck()
self.assertEqual(result.returncode, 0, result.stdout)
self.assertIn("[verdict] LOCAL_READY", result.stdout)
self.assertNotIn("[verdict] READY", result.stdout)
def test_local_precheck_rejects_main_and_missing_registration(self):
result = self.precheck(cwd=self.fixture.repo)
self.assertNotEqual(result.returncode, 0, result.stdout)
self.card.unlink()
self.save_card_change()
result = self.precheck()
self.assertNotEqual(result.returncode, 0, result.stdout)
def test_local_precheck_rejects_foreign_registration(self):
self.card.write_text(self.card.read_text().replace(str(self.wt), str(self.fixture.repo)))
self.save_card_change()
result = self.precheck()
self.assertNotEqual(result.returncode, 0, result.stdout)
def test_local_precheck_rejects_wrong_branch_registration(self):
self.card.write_text(self.card.read_text().replace("分支:feature/test", "分支:feature/other"))
self.save_card_change()
self.assertNotEqual(self.precheck().returncode, 0)
def test_local_precheck_rejects_invalid_baseline(self):
f = self.fixture
r0 = f.git(self.wt, "rev-parse", "origin/main").stdout.strip()
self.card.write_text(self.card.read_text().replace(r0, "0" * 40))
self.save_card_change()
self.assertNotEqual(self.precheck().returncode, 0)
def test_local_precheck_rejects_dirty_state_and_unknown_remote(self):
f = self.fixture
(self.wt / "tracked").write_text("preserve current edit\n")
result = self.precheck()
self.assertNotEqual(result.returncode, 0)
self.assertIn("SHARED_WORKSPACE_BUSY", result.stdout)
f.git(self.wt, "commit", "-am", "Save owner edit")
f.git(self.wt, "remote", "set-url", "origin", str(f.root / "missing.git"))
result = self.precheck()
self.assertNotEqual(result.returncode, 0)
self.assertIn("REMOTE_UNVERIFIED", result.stdout)
def test_unknown_work_mode_is_rejected(self):
result = self.precheck(mode="anything")
self.assertNotEqual(result.returncode, 0, result.stdout)
def test_local_resume_imports_original_receipt_without_sync(self):
f = self.fixture
receipt = self.capture()
original = receipt.read_bytes()
head = f.git(self.wt, "rev-parse", "HEAD").stdout
f.advance_remote()
result = self.resume(receipt)
self.assertEqual(result.returncode, 0, result.stdout)
self.assertIn("[resumed]", result.stdout)
self.assertEqual(f.git(self.wt, "rev-parse", "HEAD").stdout, head)
self.assertEqual(receipt.read_bytes(), original)
self.assertIn(json.loads(original)["receipt_sha256"], self.card.read_text())
current = f.git(self.wt, "rev-parse", "origin/main").stdout.strip()
self.assertNotEqual(json.loads(original)["after"]["target_head"], current)
def assert_history_change_rejected(self, receipt):
f = self.fixture
before = self.card.read_bytes()
original = receipt.read_bytes()
head = f.git(self.wt, "rev-parse", "HEAD").stdout
result = self.resume(receipt)
self.assertNotEqual(result.returncode, 0, result.stdout)
self.assertIn("Precheck target history has rewound or diverged", result.stdout)
self.assertNotIn("[resumed]", result.stdout)
self.assertEqual(self.card.read_bytes(), before)
self.assertEqual(receipt.read_bytes(), original)
self.assertEqual(f.git(self.wt, "rev-parse", "HEAD").stdout, head)
def test_local_resume_rejects_target_rewind_before_import(self):
f = self.fixture
r0 = f.git(self.wt, "rev-parse", "origin/main").stdout.strip()
f.advance_remote()
receipt = self.capture()
observed = json.loads(receipt.read_text())["after"]["target_head"]
self.assertNotEqual(observed, r0)
# Only this test's disposable bare remote is rewritten.
f.git(f.remote, "update-ref", "refs/heads/main", r0, observed)
# Reproduce recovery after another client operation refreshed the target.
f.git(self.wt, "fetch", "origin", "+refs/heads/main:refs/remotes/origin/main")
self.assert_history_change_rejected(receipt)
def test_local_resume_rejects_divergent_target_before_import(self):
f = self.fixture
r0 = f.git(self.wt, "rev-parse", "origin/main").stdout.strip()
f.advance_remote()
receipt = self.capture()
observed = json.loads(receipt.read_text())["after"]["target_head"]
f.git(f.repo, "switch", "--detach", r0)
(f.repo / "divergent-main").write_text("different target history\n")
f.git(f.repo, "add", "divergent-main")
f.git(f.repo, "commit", "-m", "Create divergent fixture history")
divergent = f.git(f.repo, "rev-parse", "HEAD").stdout.strip()
f.git(f.repo, "push", "origin", "HEAD:refs/heads/divergent")
self.assertEqual(f.git(f.repo, "merge-base", "--is-ancestor", observed,
divergent, check=False).returncode, 1)
# Both targets retain R0; only normal ancestry may permit recovery.
f.git(f.remote, "update-ref", "refs/heads/main", divergent, observed)
f.git(self.wt, "fetch", "origin", "+refs/heads/main:refs/remotes/origin/main")
self.assert_history_change_rejected(receipt)
def test_make_resume_exports_the_selected_mode_and_receipt(self):
f = self.fixture
receipt = self.capture()
result = f.run_cmd(self.wt, "make", "-f", str(lifecycle.SOURCE / "Makefile"),
"task-resume", check=False,
env=f.env | {"TASK_WORK_MODE": "local", "PRECHECK_RECEIPT": str(receipt)})
self.assertEqual(result.returncode, 0, result.stdout)
self.assertIn("[resumed]", result.stdout)
def test_local_receipt_cannot_start_a_task_or_resume_for_delivery(self):
f = self.fixture
receipt = self.capture()
before = self.card.read_bytes()
delivery = self.resume(receipt, mode="delivery")
self.assertNotEqual(delivery.returncode, 0, delivery.stdout)
self.assertEqual(self.card.read_bytes(), before)
start = f.task(self.wt, "start", "feature/bypass", check=False,
env=f.env | {"PRECHECK_RECEIPT": str(receipt), "TASK_WORK_MODE": "local",
"VALIDATE_RECEIPT_PURPOSE": "resume-local",
"TRIAGE_RECEIPT": str(f.triage_receipt())})
self.assertNotEqual(start.returncode, 0, start.stdout)
self.assertFalse((f.root / "repo-wt-feature-bypass").exists())
def test_strict_resume_rejects_target_advance(self):
f = self.fixture
receipt = self.capture(mode="delivery")
head = f.git(self.wt, "rev-parse", "HEAD").stdout
f.advance_remote()
rejected = self.resume(receipt, mode="delivery")
self.assertNotEqual(rejected.returncode, 0)
self.assertNotIn("[resumed]", rejected.stdout)
self.assertEqual(f.git(self.wt, "rev-parse", "HEAD").stdout, head)
def test_strict_resume_imports_current_ready_receipt(self):
receipt = self.capture(mode="delivery")
result = self.resume(receipt, mode="delivery")
self.assertEqual(result.returncode, 0, result.stdout)
def test_resume_rejects_changed_source_head(self):
receipt = self.capture()
f = self.fixture
(self.wt / "tracked").write_text("new task revision\n")
f.git(self.wt, "commit", "-am", "Change source after precheck")
before = self.card.read_bytes()
result = self.resume(receipt)
self.assertNotEqual(result.returncode, 0)
self.assertEqual(self.card.read_bytes(), before)
def test_resume_rejects_another_worktree_with_same_head(self):
f = self.fixture
other = f.root / "other-worktree"
r0 = f.git(self.wt, "rev-parse", "origin/main").stdout.strip()
other_card = self.wt / ".task/feature-other.md"
other_card.write_text(
"# Other task\n任务主记录:fixture;状态:进行中;owner:Lifecycle Test\n"
f"- 权威仓:fixture;目标分支:main;R0(initial_base_sha)= {r0}\n"
f"- 工作区:{other}(worktree);分支:feature/other\n")
f.git(self.wt, "add", ".task/feature-other.md")
f.git(self.wt, "commit", "-m", "Register second fixture task")
f.git(f.repo, "worktree", "add", "-b", "feature/other", str(other), "HEAD")
# The main checkout has an older HEAD; align this owned fixture branch to the task HEAD.
f.git(other, "merge", "--ff-only", "feature/test")
receipt = self.capture(mode="delivery")
self.assertEqual(f.git(other, "rev-parse", "HEAD").stdout,
f.git(self.wt, "rev-parse", "HEAD").stdout)
result = self.resume(receipt, mode="delivery", cwd=other)
self.assertNotEqual(result.returncode, 0)
def test_unknown_validation_purpose_is_rejected(self):
receipt = self.capture(mode="delivery")
f = self.fixture
result = f.run_cmd(self.wt, "bash", "scripts/evidence.sh", check=False,
env=f.env | {"VALIDATE_RECEIPT": str(receipt),
"VALIDATE_RECEIPT_PURPOSE": "ignore"})
self.assertNotEqual(result.returncode, 0, result.stdout)
if __name__ == "__main__":
unittest.main()