code-coordinator 0.5.192__tar.gz → 0.5.196__tar.gz

{code_coordinator-0.5.192 → code_coordinator-0.5.196}/.github/workflows/test.yml

@@ -153,6 +153,39 @@ jobs:
       # gated separately through the #2164 CI wrapper.
       - run: scripts/run_tests_without_gh.sh python -m pytest tests/ -q --ignore=tests/acceptance
 
+  # #2135 (child of epic #1902): "every `gh` argv construction lives inside
+  # coord/github_ops.py" is the load-bearing assumption under the #1902
+  # forge-portability cost estimate — it's the fact that makes the eventual
+  # port a refactor instead of a rewrite. #1483 already guarded direct
+  # call-argument leaks with an AST scan (formerly
+  # tests/test_no_direct_gh_calls.py), but that walk missed the
+  # variable-indirection shape (`cmd = ["gh", ...]; subprocess.run(cmd)`).
+  # This job's dedicated regex scan catches that gap too, so
+  # test_no_direct_gh_calls.py was retired in its favor rather than kept
+  # running alongside it as a second, disagreeing implementation of the
+  # same invariant (#2135 review) — see scripts/check_gh_argv_containment.py's
+  # module docstring for the full reconciliation.
+  #
+  # scripts/check_gh_argv_containment.py scans coord/**/*.py for a bare
+  # `"gh"`/`'gh'` token outside coord/github_ops.py, with one explicit,
+  # commented allowlist entry for coord/prereqs.py's binary-version probe
+  # (it checks `gh` is installed; it doesn't drive the forge). See that
+  # script's module docstring for the full rationale and
+  # tests/test_check_gh_argv_containment.py for the negative-test proof
+  # that the guard can actually fail.
+  #
+  # Stdlib-only (re/pathlib/dataclasses), so no `pip install` needed here —
+  # this is meant to be as cheap as `no-gh-on-path` above, a grep-shaped
+  # check, not a build step.
+  gh-argv-containment:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - run: python scripts/check_gh_argv_containment.py .
+
   # #2170: run the full suite in the environment of a REAL FLEET MACHINE — a
   # populated `~/.coord/` (thin-client shaped), no `sqlite3` on PATH, and a
   # `$TMPDIR` under an ancestor pytest config.

{code_coordinator-0.5.192 → code_coordinator-0.5.196}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-coordinator
-Version: 0.5.192
+Version: 0.5.196
 Summary: Coordinate Claude Code workers across multiple machines and repos.
 Author-email: John Donaghy <johnfdonaghy@gmail.com>
 License-Expression: LicenseRef-FSL-1.1-MIT

{code_coordinator-0.5.192 → code_coordinator-0.5.196}/code_coordinator.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-coordinator
-Version: 0.5.192
+Version: 0.5.196
 Summary: Coordinate Claude Code workers across multiple machines and repos.
 Author-email: John Donaghy <johnfdonaghy@gmail.com>
 License-Expression: LicenseRef-FSL-1.1-MIT

{code_coordinator-0.5.192 → code_coordinator-0.5.196}/code_coordinator.egg-info/SOURCES.txt

@@ -54,6 +54,7 @@ coord/client.py
 coord/comments.py
 coord/config.py
 coord/config_reload.py
+coord/confirm_test.py
 coord/conflict_fix.py
 coord/dao.py
 coord/db.py
@@ -71,6 +72,7 @@ coord/events.py
 coord/failure_class.py
 coord/filelock.py
 coord/fleet_config_health.py
+coord/forge_availability.py
 coord/freshness.py
 coord/gate_a.py
 coord/gate_b.py
@@ -307,6 +309,7 @@ docs/mocks/web/pipeline-narrow.html
 docs/mocks/web/pipeline-wide.html
 graphify-out/.gitignore
 scripts/audit_review_verdict_overrides.py
+scripts/check_gh_argv_containment.py
 scripts/codegen.py
 scripts/coord-test-runner.sh
 scripts/drive-batch.sh
@@ -377,6 +380,7 @@ tests/test_branch_model.py
 tests/test_bug_intake.py
 tests/test_build_worker_image_env_update.py
 tests/test_cargo_cache.py
+tests/test_check_gh_argv_containment.py
 tests/test_ci_acceptance_gate_1950.py
 tests/test_ci_parity_2091.py
 tests/test_ci_store.py
@@ -391,6 +395,7 @@ tests/test_cli_audit.py
 tests/test_cli_chat_continue.py
 tests/test_cli_decide.py
 tests/test_cli_diagnose_config_provenance.py
+tests/test_cli_diagnose_forge_availability.py
 tests/test_cli_diagnose_graph.py
 tests/test_cli_diagnose_self.py
 tests/test_cli_doctor.py
@@ -449,6 +454,7 @@ tests/test_config_audit.py
 tests/test_config_pricing.py
 tests/test_config_snapshot_noncanonical_2208.py
 tests/test_config_usage_gate.py
+tests/test_confirm_test.py
 tests/test_conflict_fix.py
 tests/test_conflict_fix_semantic.py
 tests/test_coord_test.py
@@ -494,6 +500,7 @@ tests/test_fix_of_force_override.py
 tests/test_fleet_config_health.py
 tests/test_fleet_health_probes.py
 tests/test_fleet_health_snapshot.py
+tests/test_forge_availability.py
 tests/test_freshness.py
 tests/test_gate_a.py
 tests/test_gate_b.py
@@ -549,7 +556,6 @@ tests/test_needs_attention.py
 tests/test_network.py
 tests/test_new_issue_chat.py
 tests/test_next_release_tag.py
-tests/test_no_direct_gh_calls.py
 tests/test_node_shim.py
 tests/test_notifier_baseline.py
 tests/test_notifier_cli.py

{code_coordinator-0.5.192 → code_coordinator-0.5.196}/code_coordinator.egg-info/scm_file_list.json

@@ -78,6 +78,7 @@
     "coord/comments.py",
     "coord/config.py",
     "coord/config_reload.py",
+    "coord/confirm_test.py",
     "coord/conflict_fix.py",
     "coord/dao.py",
     "coord/dashboard/__init__.py",
@@ -117,6 +118,7 @@
     "coord/failure_class.py",
     "coord/filelock.py",
     "coord/fleet_config_health.py",
+    "coord/forge_availability.py",
     "coord/freshness.py",
     "coord/gate_a.py",
     "coord/gate_b.py",
@@ -311,6 +313,7 @@
     "scripts/azure-workers/scrub-and-generalize.sh",
     "scripts/azure-workers/tailnet-acl.hujson",
     "scripts/azure-workers/verify-github-token.sh",
+    "scripts/check_gh_argv_containment.py",
     "scripts/codegen.py",
     "scripts/coord-test-runner.sh",
     "scripts/drive-batch.sh",
@@ -439,6 +442,7 @@
     "tests/test_bug_intake.py",
     "tests/test_build_worker_image_env_update.py",
     "tests/test_cargo_cache.py",
+    "tests/test_check_gh_argv_containment.py",
     "tests/test_ci_acceptance_gate_1950.py",
     "tests/test_ci_parity_2091.py",
     "tests/test_ci_store.py",
@@ -453,6 +457,7 @@
     "tests/test_cli_chat_continue.py",
     "tests/test_cli_decide.py",
     "tests/test_cli_diagnose_config_provenance.py",
+    "tests/test_cli_diagnose_forge_availability.py",
     "tests/test_cli_diagnose_graph.py",
     "tests/test_cli_diagnose_self.py",
     "tests/test_cli_doctor.py",
@@ -511,6 +516,7 @@
     "tests/test_config_pricing.py",
     "tests/test_config_snapshot_noncanonical_2208.py",
     "tests/test_config_usage_gate.py",
+    "tests/test_confirm_test.py",
     "tests/test_conflict_fix.py",
     "tests/test_conflict_fix_semantic.py",
     "tests/test_coord_test.py",
@@ -556,6 +562,7 @@
     "tests/test_fleet_config_health.py",
     "tests/test_fleet_health_probes.py",
     "tests/test_fleet_health_snapshot.py",
+    "tests/test_forge_availability.py",
     "tests/test_freshness.py",
     "tests/test_gate_a.py",
     "tests/test_gate_b.py",
@@ -611,7 +618,6 @@
     "tests/test_network.py",
     "tests/test_new_issue_chat.py",
     "tests/test_next_release_tag.py",
-    "tests/test_no_direct_gh_calls.py",
     "tests/test_node_shim.py",
     "tests/test_notifier_baseline.py",
     "tests/test_notifier_cli.py",

code_coordinator-0.5.196/code_coordinator.egg-info/scm_version.json

@@ -0,0 +1,8 @@
+{
+  "tag": "0.5.196",
+  "distance": 0,
+  "node": "g049ba86f1fa5713231873ab0a75ea2fa2c2fe27f",
+  "dirty": false,
+  "branch": "HEAD",
+  "node_date": "2026-08-21"
+}

{code_coordinator-0.5.192 → code_coordinator-0.5.196}/coord/ci_github.py

@@ -14,11 +14,13 @@ from __future__ import annotations
 import re
 import subprocess
 import time
+from collections import Counter
 from dataclasses import dataclass, field
 from datetime import datetime
 
 from coord import github_ops
 from coord.ci_store import CheckRun, JobRun, JobStep, failed_checks
+from coord.forge_availability import record_ci_check_fetch
 
 
 def _parse_ts(raw: str | None) -> float | None:
@@ -381,6 +383,18 @@ class GitHubCi:
     # ── Internal ────────────────────────────────────────────────────────────
 
     def _fetch(self, repo: str, number: int) -> list[CheckRun]:
+        # #1896 Phase 0: this is the "every CiStore.list_checks_for_pr
+        # outcome" seam the forge-availability program asks for — reachable/
+        # unreachable, plus the check-level conclusion distribution.
+        # `github_ops.get_pr_checks` does NOT go through `_gh()` — it shells
+        # out to `gh pr checks` directly (a non-zero exit there can still
+        # carry usable JSON on stdout, which `_gh()`'s raise-on-nonzero
+        # contract can't express, see `get_pr_checks`'s own docstring) — so
+        # this call is the *only* forge-availability observation this read
+        # produces. It stands alone, not "layered on top" of anything.
+        # Only fires on a real cache miss — a cached `list_checks_for_pr`
+        # hit never reaches here, so this costs nothing extra either.
+        _t0 = time.monotonic()
         try:
             raw = github_ops.get_pr_checks(repo, number)
         except github_ops.GhTooOldForJsonChecks as e:
@@ -394,6 +408,8 @@ class GitHubCi:
             # it into `_unreadable_check`'s generic wording) means an operator
             # reading the merge refusal never has to guess which of the two
             # this was.
+            record_ci_check_fetch(repo, number, outcome="unreachable",
+                                   duration_s=time.monotonic() - _t0, detail="gh too old")
             return [_gh_too_old_check(repo, number, str(e))]
         except (FileNotFoundError, subprocess.TimeoutExpired, RuntimeError, ValueError) as e:
             # #1525: a `gh pr checks` read that outright failed (gh missing,
@@ -407,8 +423,13 @@ class GitHubCi:
             # failing check instead so the gate blocks and says why; a caller
             # that genuinely wants "unknown" treated as clear must pass
             # `force_merge=True` explicitly.
+            record_ci_check_fetch(repo, number, outcome="unreachable",
+                                   duration_s=time.monotonic() - _t0, detail=str(e))
             return [_unreadable_check(repo, number, str(e))]
+        duration = time.monotonic() - _t0
         if not isinstance(raw, list):
+            record_ci_check_fetch(repo, number, outcome="unreachable",
+                                   duration_s=duration, detail="non-list JSON")
             return [_unreadable_check(repo, number, "gh pr checks returned non-list JSON")]
         # #2446: `_all_checks` (this method's only caller) is the single
         # unfiltered fetch shared by both `list_checks_for_pr` (narrowed to
@@ -416,7 +437,7 @@ class GitHubCi:
         # required-contexts narrowing used to happen here, which meant an
         # advisory check's regression was invisible everywhere, not just to
         # the merge gate. See both methods' docstrings.
-        return [
+        checks = [
             CheckRun(
                 name=str(entry.get("name", "")),
                 status=_status_from_bucket(str(entry.get("bucket", ""))),
@@ -429,6 +450,10 @@ class GitHubCi:
             for entry in raw
             if isinstance(entry, dict)
         ]
+        conclusions = Counter(c.conclusion or "pending" for c in checks)
+        record_ci_check_fetch(repo, number, outcome="ok", duration_s=duration,
+                               conclusions=dict(conclusions))
+        return checks
 
     def _required_contexts(self, repo: str) -> frozenset[str] | None:
         now = time.time()

{code_coordinator-0.5.192 → code_coordinator-0.5.196}/coord/commands/lifecycle.py

@@ -95,9 +95,23 @@ def notify(config_path: Path) -> None:
     _svc = daemon_reroute_target("COORD_NOTIFY_ON_DAEMON")
     if _svc is not None:
         from coord.client import post_record  # noqa: PLC0415
+        from coord.confirm_test import (  # noqa: PLC0415
+            notify_client_timeout_seconds,
+        )
 
         try:
-            resp = post_record(_svc, "/notify", {}, timeout=180.0)
+            # #2464: a drain now re-runs the repo's real build+test to confirm
+            # Test-stage PASS claims, so a pass can legitimately outlast the
+            # pre-#2464 180s.  The daemon runs it to completion under
+            # `notify.lock` regardless of what this client does (see
+            # `serve_app.post_notify`), so giving up early would report
+            # "notify via daemon failed" for a pass that is still running and
+            # will finish fine — on the one command the auto-loop is driven
+            # by.  Same fix, same shape, as `coord merge --revalidate`'s
+            # `coord.revalidate.client_timeout_seconds` (#1769/#1715).
+            resp = post_record(
+                _svc, "/notify", {}, timeout=notify_client_timeout_seconds(),
+            )
         except Exception as exc:  # noqa: BLE001
             click.echo(f"error: notify via daemon failed: {exc}", err=True)
             sys.exit(1)

{code_coordinator-0.5.192 → code_coordinator-0.5.196}/coord/commands/status.py

@@ -1585,6 +1585,26 @@ def _diagnose_via_daemon(svc, params: dict) -> None:
         "behind."
     ),
 )
+@click.option(
+    "--forge-availability",
+    "forge_availability",
+    is_flag=True,
+    help=(
+        "#1896: report forge/CI availability over a trailing window — "
+        "uptime %, longest contiguous unavailable stretch, and merge-gate "
+        "CI refusal counts by reason — from observations recorded at the "
+        "gh/CI/merge-gate seams. Read-only, local audit_log only. Pair "
+        "with --window-days to change the window (default 30)."
+    ),
+)
+@click.option(
+    "--window-days",
+    "forge_availability_window_days",
+    default=30.0,
+    show_default=True,
+    type=float,
+    help="With --forge-availability, the trailing window (days) to summarize.",
+)
 
 
 @_CONFIG_OPTION
@@ -1601,6 +1621,8 @@ def diagnose(
     config_provenance_check: bool = False,
     self_check: bool = False,
     self_no_fetch: bool = False,
+    forge_availability: bool = False,
+    forge_availability_window_days: float = 30.0,
 ) -> None:
     """Per-stage doctor — diagnose, best-effort recover, optional reset."""
     # ── graphify graph freshness sweep (read-only) ───────────────────────────
@@ -1608,6 +1630,11 @@ def diagnose(
         _diagnose_graph_health(config_path)
         return
 
+    # ── #1896: forge/CI availability read-out (read-only) ───────────────────
+    if forge_availability:
+        _diagnose_forge_availability(forge_availability_window_days)
+        return
+
     # ── #1779: fleet coordinator.yml provenance (read-only) ─────────────────
     if config_provenance_check:
         _diagnose_config_provenance()
@@ -1762,6 +1789,28 @@ def _diagnose_graph_health(config_path: Path) -> None:
     )
 
 
+def _diagnose_forge_availability(window_days: float) -> None:
+    """Report forge/CI availability over the trailing *window_days* (#1896).
+
+    Read-only: summarizes the ``forge_availability``-category rows in the
+    local ``audit_log`` (recorded best-effort at the ``gh`` call seam, the
+    CI check-fetch seam, and the merge-gate refusal seam — see
+    :mod:`coord.forge_availability`). Same family as ``--graph``/
+    ``--config-provenance``/``--self``: local-machine only, no network call
+    of its own.
+    """
+    from coord.forge_availability import (  # noqa: PLC0415
+        availability_report,
+        format_report_lines,
+        summary_line as forge_summary_line,
+    )
+
+    report = availability_report(window_days=window_days)
+    for line in format_report_lines(report):
+        click.echo(f"  {line}")
+    click.echo(forge_summary_line(report))
+
+
 def _diagnose_config_provenance() -> None:
     """Report whether THIS machine's live ``coordinator.yml`` is still the
     reviewed one (#1779).

{code_coordinator-0.5.192 → code_coordinator-0.5.196}/coord/config.py

@@ -977,6 +977,18 @@ class PipelineConfig:
     per-turn liveness auditor — see :class:`LivenessAuditorConfig`. Off by
     default; gates nothing even when enabled (see
     :func:`coord.notify.detect_liveness_stall`).
+
+    ``confirm_test_verdict`` (#2464) controls whether a Test-stage PASS
+    claim is independently re-run before it is recorded — see
+    :mod:`coord.confirm_test`. **Defaults to ``True``**, unlike the
+    ships-dark flags above, and deliberately so: the thing it guards is a
+    verdict the worker issued about its own work, and a correctness gate
+    that must be switched on is the posture that let the defect ship. It is
+    also cheap to leave on where it cannot apply — a machine with no local
+    checkout of the repo returns "inconclusive" and the pre-#2464 behaviour
+    stands. Costs one real suite run per Test-stage completion where it
+    *can* apply; set ``false`` (or ``COORD_CONFIRM_TEST_VERDICT=0``) to opt
+    back out.
     """
 
     default_gates: list[str] = field(default_factory=lambda: ["test", "review", "merge"])
@@ -995,6 +1007,8 @@ class PipelineConfig:
     convergence_rounds: int = 3
     # #2048 — DEFAULT OFF. See LivenessAuditorConfig's docstring.
     liveness_auditor: LivenessAuditorConfig = field(default_factory=LivenessAuditorConfig)
+    # #2464 — DEFAULT ON. See the class docstring.
+    confirm_test_verdict: bool = True
 
     def attention_threshold_for(
         self,
@@ -2697,6 +2711,12 @@ def _parse_pipeline(raw: Any) -> PipelineConfig:
             raise ConfigError("pipeline.auto_loop must be a boolean")
         cfg.auto_loop = value
 
+    if "confirm_test_verdict" in raw:
+        value = raw["confirm_test_verdict"]
+        if not isinstance(value, bool):
+            raise ConfigError("pipeline.confirm_test_verdict must be a boolean")
+        cfg.confirm_test_verdict = value
+
     if "max_review_iterations" in raw:
         value = raw["max_review_iterations"]
         if not isinstance(value, int) or value < 1: