test: reorganize suite into unit/integration/canaries directories

Replace the hand-maintained INTEGRATION_NAMES classifier (and the
bespoke run_tests.py around it) with a directory-driven split:

  tests/unit/         unit tests, always run
  tests/integration/  Docker-dependent, skip cleanly without Docker
  tests/canaries/     upstream-regression checks, opt-in via
                      CLAUDE_BOTTLE_RUN_CANARIES=1

The pinned-pipelock-image check moves to the canary suite — it tests
upstream packaging, not our code, so it shouldn't gate every dev push.
A scheduled canaries.yml workflow runs it weekly.

The manifest-runtime tests collapse the four assertRaises cases for
distinct 'runtime' values into one subTest loop and drop the
error-message-wording assertions; the contract is "any value is
rejected", not "the error literally contains 'auto-detect'".

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

tests/integration/test_dry_run_plan.py

@@ -0,0 +1,81 @@
+"""Integration: cli.py start --dry-run renders the planned shape and
+does not create any docker resources. Confirms the preflight contract
+from PRD 0001 (allowlist line in the plan, no docker side effects)."""
+
+import json
+import os
+import re
+import subprocess
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+
+from tests._docker import skip_unless_docker
+
+REPO_ROOT = Path(__file__).resolve().parent.parent
+
+
+@skip_unless_docker()
+class TestDryRunPlan(unittest.TestCase):
+    def test_dry_run(self):
+        work_dir = Path(tempfile.mkdtemp())
+        try:
+            manifest = work_dir / "claude-bottle.json"
+            manifest.write_text(json.dumps({
+                "bottles": {"dev": {"egress": {"allowlist": ["example.org"]}}},
+                "agents": {
+                    "demo": {"skills": [], "prompt": "", "bottle": "dev"},
+                },
+            }))
+
+            nets_before = self._count_claude_bottle_networks()
+            ctrs_before = self._count_claude_bottle_containers()
+
+            env = os.environ.copy()
+            env["HOME"] = str(work_dir)
+            env["CLAUDE_BOTTLE_DRY_RUN"] = "1"
+            result = subprocess.run(
+                [sys.executable, str(REPO_ROOT / "cli.py"), "start", "demo"],
+                cwd=work_dir,
+                env=env,
+                capture_output=True,
+                text=True,
+            )
+            out = result.stdout + result.stderr
+
+            self.assertIn("egress", out, "preflight: egress line present")
+            # 7 baked defaults + 1 bottle entry = 8.
+            self.assertRegex(out, r"8 hosts allowed", "preflight: bottle entry counted")
+            self.assertIn("api.anthropic.com", out, "preflight: baked default shown")
+            self.assertRegex(out, r"runtime\s*:\s*runc", "preflight: default runtime shown")
+            self.assertIn("dry-run requested", out, "dry-run banner present")
+            self.assertNotIn("/dev/tty", out, "dry-run exited before tty prompt")
+
+            self.assertEqual(nets_before, self._count_claude_bottle_networks(),
+                             "no networks created")
+            self.assertEqual(ctrs_before, self._count_claude_bottle_containers(),
+                             "no containers created")
+        finally:
+            import shutil
+            shutil.rmtree(work_dir, ignore_errors=True)
+
+    def _count_claude_bottle_networks(self) -> int:
+        result = subprocess.run(
+            ["docker", "network", "ls", "--format", "{{.Name}}"],
+            capture_output=True,
+            text=True,
+        )
+        return sum(1 for n in result.stdout.splitlines() if n.startswith("claude-bottle"))
+
+    def _count_claude_bottle_containers(self) -> int:
+        result = subprocess.run(
+            ["docker", "ps", "-a", "--format", "{{.Names}}"],
+            capture_output=True,
+            text=True,
+        )
+        return sum(1 for n in result.stdout.splitlines() if n.startswith("claude-bottle"))
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/integration/test_orphan_cleanup.py

@@ -0,0 +1,79 @@
+"""Integration: the cleanup primitives the start-flow trap depends on
+are idempotent. The original orphan-network bug was a trap-ordering
+issue; the fix moved the install earlier. The trap is only safe if
+network_remove and PipelockProxy.stop are no-ops against missing
+resources."""
+
+import os
+import subprocess
+import unittest
+
+from claude_bottle.backend.docker.network import (
+    network_create_egress,
+    network_create_internal,
+    network_remove,
+)
+from claude_bottle.backend.docker.pipelock import (
+    DockerPipelockProxy,
+    pipelock_container_name,
+)
+from tests._docker import skip_unless_docker
+
+
+@skip_unless_docker()
+class TestOrphanCleanup(unittest.TestCase):
+    def setUp(self):
+        self.slug = f"cb-test-orphan-{os.getpid()}"
+        self.internal_name = ""
+        self.egress_name = ""
+
+    def tearDown(self):
+        for n in (self.internal_name, self.egress_name):
+            if n:
+                subprocess.run(
+                    ["docker", "network", "rm", n],
+                    stdout=subprocess.DEVNULL,
+                    stderr=subprocess.DEVNULL,
+                )
+
+    def test_remove_missing_is_noop(self):
+        # Returning True == idempotent success.
+        self.assertTrue(network_remove(f"claude-bottle-net-{self.slug}-does-not-exist"))
+
+    @unittest.skipIf(
+        os.environ.get("GITEA_ACTIONS") == "true",
+        "skipped under act_runner: docker socket mount topology breaks "
+        "in-process visibility of networks created on the host daemon",
+    )
+    def test_create_and_remove(self):
+        self.internal_name = network_create_internal(self.slug)
+        self.egress_name = network_create_egress(self.slug)
+
+        nets = subprocess.run(
+            ["docker", "network", "ls", "--format", "{{.Name}}"],
+            capture_output=True, text=True,
+        ).stdout.splitlines()
+        self.assertIn(self.internal_name, nets)
+        self.assertIn(self.egress_name, nets)
+
+        self.assertTrue(network_remove(self.internal_name))
+        self.assertTrue(network_remove(self.egress_name))
+
+        nets_after = subprocess.run(
+            ["docker", "network", "ls", "--format", "{{.Name}}"],
+            capture_output=True, text=True,
+        ).stdout.splitlines()
+        self.assertNotIn(self.internal_name, nets_after)
+        self.assertNotIn(self.egress_name, nets_after)
+
+        # Idempotent on already-removed.
+        self.assertTrue(network_remove(self.internal_name))
+        self.assertTrue(network_remove(self.egress_name))
+
+    def test_pipelock_stop_missing_sidecar(self):
+        # Should not raise.
+        DockerPipelockProxy().stop(pipelock_container_name(f"missing-{self.slug}"))
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/integration/test_pipelock_sidecar_smoke.py

@@ -0,0 +1,104 @@
+"""Integration: full sidecar smoke test. Boots a pipelock container the
+same way cli.py does (docker create + docker cp YAML + docker start),
+then probes /health."""
+
+import os
+import re
+import shutil
+import subprocess
+import tempfile
+import time
+import unittest
+import urllib.request
+from pathlib import Path
+
+from claude_bottle.backend.docker.pipelock import (
+    PIPELOCK_IMAGE,
+    DockerPipelockProxy,
+)
+from tests._docker import skip_unless_docker
+from tests.fixtures import fixture_minimal
+
+
+@skip_unless_docker()
+class TestPipelockSidecarSmoke(unittest.TestCase):
+    def setUp(self):
+        self.name = f"cb-test-pipelock-smoke-{os.getpid()}"
+        self.work_dir = Path(tempfile.mkdtemp())
+
+    def tearDown(self):
+        subprocess.run(
+            ["docker", "rm", "-f", self.name],
+            stdout=subprocess.DEVNULL,
+            stderr=subprocess.DEVNULL,
+        )
+        shutil.rmtree(self.work_dir, ignore_errors=True)
+
+    @unittest.skipIf(
+        os.environ.get("GITEA_ACTIONS") == "true",
+        "skipped under act_runner: published port is on the host's "
+        "loopback, not reachable from the job container's 127.0.0.1",
+    )
+    def test_smoke(self):
+        yaml_path = self.work_dir / "pipelock.yaml"
+        DockerPipelockProxy().prepare(fixture_minimal().bottles["dev"], "demo", yaml_path)
+
+        create = subprocess.run(
+            [
+                "docker", "create",
+                "--name", self.name,
+                "-p", "0:8888",
+                PIPELOCK_IMAGE,
+                "run", "--config", "/etc/pipelock.yaml",
+                "--listen", "0.0.0.0:8888",
+            ],
+            stdout=subprocess.DEVNULL,
+            stderr=subprocess.PIPE,
+            text=True,
+        )
+        self.assertEqual(0, create.returncode, f"docker create failed: {create.stderr}")
+
+        # Guard against /etc/pipelock/ regressions: the path must be
+        # /etc/pipelock.yaml, since the image is distroless.
+        cp = subprocess.run(
+            ["docker", "cp", str(yaml_path), f"{self.name}:/etc/pipelock.yaml"],
+            stdout=subprocess.DEVNULL,
+            stderr=subprocess.PIPE,
+            text=True,
+        )
+        self.assertEqual(0, cp.returncode, f"docker cp failed: {cp.stderr}")
+
+        start = subprocess.run(
+            ["docker", "start", self.name],
+            stdout=subprocess.DEVNULL,
+            stderr=subprocess.PIPE,
+            text=True,
+        )
+        self.assertEqual(0, start.returncode,
+                         f"docker start failed; check argv 'run --listen 0.0.0.0:8888'")
+
+        port_result = subprocess.run(
+            ["docker", "port", self.name, "8888"],
+            capture_output=True, text=True,
+        )
+        first_line = (port_result.stdout or "").splitlines()[0] if port_result.stdout else ""
+        host_port = first_line.rsplit(":", 1)[-1] if first_line else ""
+        self.assertTrue(host_port, "could not determine published port")
+
+        health_url = f"http://127.0.0.1:{host_port}/health"
+        body = ""
+        for _ in range(15):
+            try:
+                with urllib.request.urlopen(health_url, timeout=2) as resp:
+                    body = resp.read().decode("utf-8")
+                    break
+            except (urllib.error.URLError, urllib.error.HTTPError, ConnectionError):
+                time.sleep(1)
+
+        self.assertIn('"status":"healthy"', body, "health body status:healthy")
+        self.assertRegex(body, r'"version":"[0-9]+\.[0-9]+\.[0-9]+"',
+                         "health body has version field")
+
+
+if __name__ == "__main__":
+    unittest.main()