Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

fix(change-risk): harden get_change_risk subprocess handling and align with MCP tool conventions (#899)

Raghav Chamadiyaffd2797ecedd
diff --git a/packages/core/src/repowise/core/analysis/change_risk/baseline.py b/packages/core/src/repowise/core/analysis/change_risk/baseline.py
index 9c43441b4..6a2830092 100644
--- a/packages/core/src/repowise/core/analysis/change_risk/baseline.py
+++ b/packages/core/src/repowise/core/analysis/change_risk/baseline.py
@@ -12,7 +12,7 @@ import subprocess
import pathspec
-from .features import features_from_file_changes
+from .features import GIT_TIMEOUT_SECONDS, features_from_file_changes
from .model import score_change
@@ -35,6 +35,10 @@ def baseline_scores(
*exclude_patterns* use gitignore syntax and are applied to every sampled
commit, matching the target change's filtering.
"""
+ # stdin=DEVNULL + timeout: a stuck git must not hang the caller (on MCP
+ # stdio transport an inherited pipe handle can wedge the whole session).
+ # No returncode check: the anchor was already validated by the feature
+ # extraction, and a failed sample degrades honestly to "no percentile".
out = subprocess.run(
["git", "log", f"-n{limit}", "--no-merges", "--format=%x1e%H", "--numstat", anchor],
cwd=repo_path,
@@ -42,6 +46,8 @@ def baseline_scores(
text=True,
encoding="utf-8",
errors="replace",
+ stdin=subprocess.DEVNULL,
+ timeout=GIT_TIMEOUT_SECONDS,
).stdout
scores: list[float] = []
diff --git a/packages/core/src/repowise/core/analysis/change_risk/features.py b/packages/core/src/repowise/core/analysis/change_risk/features.py
index cd1ab6689..c3f7fa504 100644
--- a/packages/core/src/repowise/core/analysis/change_risk/features.py
+++ b/packages/core/src/repowise/core/analysis/change_risk/features.py
@@ -47,15 +47,32 @@ class ChangeFeatures:
ref: str = "" # the commit sha or "base..head" range scored
-def _git(args: list[str], cwd: str) -> str:
- return subprocess.run(
+# Generous ceiling: even a 200-commit numstat walk finishes in seconds. The
+# point is that a stuck git (lock contention, network filesystem) must fail
+# loud instead of hanging the caller's thread forever.
+GIT_TIMEOUT_SECONDS = 60
+
+
+def _git(args: list[str], cwd: str, *, check: bool = True) -> str:
+ # stdin=DEVNULL: on MCP stdio transport a child that inherits the JSON-RPC
+ # pipe handles can wedge the session (same failure mode _meta.py guards
+ # against). check=True so a bad revspec raises instead of yielding empty
+ # stdout, which used to score as a zero-feature "low risk" change.
+ proc = subprocess.run(
["git", *args],
cwd=cwd,
capture_output=True,
text=True,
encoding="utf-8",
errors="replace",
- ).stdout
+ stdin=subprocess.DEVNULL,
+ timeout=GIT_TIMEOUT_SECONDS,
+ )
+ if check and proc.returncode != 0:
+ raise subprocess.CalledProcessError(
+ proc.returncode, proc.args, output=proc.stdout, stderr=proc.stderr
+ )
+ return proc.stdout
def _accumulate_numstat(
@@ -103,9 +120,12 @@ def _author_experience(repo_path: str, author: str, upto_ref: str) -> int:
"""Author's prior commit count reachable from *upto_ref* (one cheap call)."""
if not author:
return 0
+ # check=False: --author is a regex, so a name with metacharacters can make
+ # git error; unknown experience degrades to 0 rather than failing the score.
out = _git(
["rev-list", "--count", "--author", author, "--no-merges", upto_ref],
repo_path,
+ check=False,
).strip()
try:
return int(out)
@@ -178,7 +198,10 @@ def extract_commit_features(
author, _, subject = meta.partition("\x00")
numstat = _git(["show", sha, "--numstat", "--format="], repo_path)
la, ld, nf, dirs, subs, per_file = _accumulate_numstat(numstat, extensions, exclude_patterns)
- parent = _git(["rev-parse", "--verify", "--quiet", f"{sha}^"], repo_path).strip()
+ # check=False: a root commit has no parent and that is not an error.
+ parent = _git(
+ ["rev-parse", "--verify", "--quiet", f"{sha}^"], repo_path, check=False
+ ).strip()
exp = _author_experience(repo_path, author, parent or sha)
return ChangeFeatures(
la=la,
diff --git a/packages/core/src/repowise/core/analysis/change_risk/service.py b/packages/core/src/repowise/core/analysis/change_risk/service.py
index 48821b6af..3d9e4a4cc 100644
--- a/packages/core/src/repowise/core/analysis/change_risk/service.py
+++ b/packages/core/src/repowise/core/analysis/change_risk/service.py
@@ -7,7 +7,12 @@ from dataclasses import dataclass, replace
from pathlib import Path
from .baseline import baseline_scores
-from .features import ChangeFeatures, extract_commit_features, extract_range_features
+from .features import (
+ GIT_TIMEOUT_SECONDS,
+ ChangeFeatures,
+ extract_commit_features,
+ extract_range_features,
+)
from .model import ChangeRisk, score_change
from .normalize import RiskNormalizer, review_priority_classification
@@ -34,6 +39,8 @@ def riskignore_patterns(repo_path: str) -> tuple[str, ...]:
capture_output=True,
text=True,
encoding="utf-8",
+ stdin=subprocess.DEVNULL,
+ timeout=GIT_TIMEOUT_SECONDS,
)
if proc.returncode != 0:
return ()
diff --git a/packages/server/src/repowise/server/mcp_server/tool_change_risk.py b/packages/server/src/repowise/server/mcp_server/tool_change_risk.py
index e2720fb81..50b7d58de 100644
--- a/packages/server/src/repowise/server/mcp_server/tool_change_risk.py
+++ b/packages/server/src/repowise/server/mcp_server/tool_change_risk.py
@@ -4,10 +4,12 @@ from __future__ import annotations
import asyncio
import subprocess
+import time
from repowise.core.analysis.change_risk import change_risk_payload, score_live_change
from repowise.core.registry import mcp_tool_registry as mcp
-from repowise.server.mcp_server._helpers import _resolve_repo_context
+from repowise.server.mcp_server._helpers import _resolve_repo_context, _unsupported_repo_all
+from repowise.server.mcp_server._meta import build_meta as _build_meta
@mcp.tool()
@@ -41,7 +43,10 @@ async def get_change_risk(
exclude_patterns: Gitignore-style paths to omit, for example ``["tests/", "*.md"]``.
baseline: Recent commits to sample for percentile ranking; 0 disables it.
"""
+ if repo == "all":
+ return _unsupported_repo_all("get_change_risk")
ctx = await _resolve_repo_context(repo)
+ started = time.perf_counter()
try:
result = await asyncio.to_thread(
score_live_change,
@@ -54,5 +59,20 @@ async def get_change_risk(
except ValueError as exc:
return {"error": str(exc)}
except subprocess.CalledProcessError as exc:
- return {"error": f"Could not read change {revspec!r}: {exc.stderr or exc}"}
- return change_risk_payload(result)
+ detail = (exc.stderr or "").strip() or str(exc)
+ return {"error": f"Could not read change {revspec!r}: {detail}"}
+ except subprocess.TimeoutExpired:
+ return {"error": f"git timed out reading change {revspec!r}."}
+ payload = change_risk_payload(result)
+ if result.features.nf == 0:
+ payload["warning"] = (
+ f"No counted file changes in {revspec!r} "
+ "(check the revspec, extensions, or exclusion filters)."
+ )
+ # source: live_git marks that this response is computed from the working
+ # checkout's git, not the index, so index freshness does not apply to it.
+ payload["_meta"] = _build_meta(
+ timing_ms=(time.perf_counter() - started) * 1000,
+ extra={"source": "live_git"},
+ )
+ return payload
diff --git a/packages/server/src/repowise/server/mcp_server/tool_risk/get_risk.py b/packages/server/src/repowise/server/mcp_server/tool_risk/get_risk.py
index cbc2ec699..996e1eeb5 100644
--- a/packages/server/src/repowise/server/mcp_server/tool_risk/get_risk.py
+++ b/packages/server/src/repowise/server/mcp_server/tool_risk/get_risk.py
@@ -44,6 +44,8 @@ async def get_risk(
(will_break, missing_cochanges, missing_tests, tests_to_run) — read it
first. tests_to_run is coverage-backed: the tests the per-test map proves
exercise the changed files, empty when no coverage map is ingested.
+ To score a live commit or ``base..head`` diff by revspec instead of file
+ paths, use ``get_change_risk``.
Args:
targets: file paths to assess.
diff --git a/tests/unit/analysis/test_change_risk.py b/tests/unit/analysis/test_change_risk.py
index a941092de..24166087b 100644
--- a/tests/unit/analysis/test_change_risk.py
+++ b/tests/unit/analysis/test_change_risk.py
@@ -16,6 +16,7 @@ from repowise.core.analysis.change_risk import (
extract_range_features,
features_from_file_changes,
score_change,
+ score_live_change,
)
from repowise.core.analysis.change_risk.model import _CONSTANTS, _sigmoid
@@ -278,3 +279,42 @@ def test_score_change_on_real_commit(git_repo: Path) -> None:
assert 0.0 <= risk.score <= 10.0
assert risk.features is f
assert not math.isnan(risk.probability)
+
+
+def test_extract_commit_features_bad_revspec_raises(git_repo: Path) -> None:
+ # A bogus revspec must raise (git returns nonzero), not silently produce an
+ # all-zero feature vector that scores as a low-risk change.
+ with pytest.raises(subprocess.CalledProcessError):
+ extract_commit_features(str(git_repo), "no-such-ref")
+
+
+def test_score_live_change_bad_revspec_raises(git_repo: Path) -> None:
+ with pytest.raises(subprocess.CalledProcessError):
+ score_live_change(str(git_repo), "no-such-ref", baseline=0)
+
+
+def test_score_live_change_rejects_negative_baseline(git_repo: Path) -> None:
+ with pytest.raises(ValueError, match="baseline"):
+ score_live_change(str(git_repo), "HEAD", baseline=-1)
+
+
+def test_score_live_change_three_dot_range_is_valid(git_repo: Path) -> None:
+ base = subprocess.run(
+ ["git", "rev-parse", "HEAD"], cwd=git_repo, check=True, capture_output=True, text=True
+ ).stdout.strip()
+ _commit(git_repo, {"r/a.py": "a=1\n"}, "feat: a", author="Dev")
+ # Three-dot syntax (base...HEAD) must degrade to a valid anchor rather than
+ # leaving a leading dot that git rejects as a ref.
+ result = score_live_change(str(git_repo), f"{base}...HEAD", baseline=0)
+ assert result.features.nf == 1
+ assert result.features.ref == f"{base}..HEAD"
+
+
+def test_score_live_change_below_min_baseline_yields_no_percentile(git_repo: Path) -> None:
+ # A shallow repo (fewer than the 8-commit floor) can't produce a
+ # representative percentile, so it degrades to no ranking, not a wrong one.
+ _commit(git_repo, {"a.py": "a=1\n"}, "feat: a", author="Dev")
+ result = score_live_change(str(git_repo), "HEAD", baseline=200)
+ assert result.baseline_sample_size < 8
+ assert result.percentile is None
+ assert result.priority is None
diff --git a/tests/unit/server/mcp/test_change_risk.py b/tests/unit/server/mcp/test_change_risk.py
index 77d8b02bf..ffda42c71 100644
--- a/tests/unit/server/mcp/test_change_risk.py
+++ b/tests/unit/server/mcp/test_change_risk.py
@@ -64,3 +64,58 @@ async def test_get_change_risk_honors_riskignore_and_request_filters(tmp_path, m
assert result["review_priority"] is None
assert result["classification"] is None
assert result["baseline_sample_size"] == 0
+ # Live-git responses carry a _meta envelope flagged as index-independent.
+ assert result["_meta"]["source"] == "live_git"
+ assert "warning" not in result
+
+
+async def test_get_change_risk_bad_revspec_returns_error(tmp_path, monkeypatch) -> None:
+ repo = tmp_path / "repo"
+ repo.mkdir()
+ _git(["init", "-q"], repo)
+ _commit(repo, {"README.md": "# seed\n"}, "chore: seed")
+
+ module = importlib.import_module("repowise.server.mcp_server.tool_change_risk")
+
+ async def _context(_: str | None) -> SimpleNamespace:
+ return SimpleNamespace(path=str(repo))
+
+ monkeypatch.setattr(module, "_resolve_repo_context", _context)
+ result = await module.get_change_risk(revspec="does-not-exist", baseline=0)
+
+ # A bogus revspec must surface an error, not a silent zero-risk score.
+ assert "error" in result
+ assert "does-not-exist" in result["error"]
+ assert "score" not in result
+
+
+async def test_get_change_risk_empty_diff_warns(tmp_path, monkeypatch) -> None:
+ repo = tmp_path / "repo"
+ repo.mkdir()
+ _git(["init", "-q"], repo)
+ _commit(repo, {"README.md": "# seed\n"}, "chore: seed")
+ _commit(repo, {"app.py": "value = 1\n"}, "feat: app")
+
+ module = importlib.import_module("repowise.server.mcp_server.tool_change_risk")
+
+ async def _context(_: str | None) -> SimpleNamespace:
+ return SimpleNamespace(path=str(repo))
+
+ monkeypatch.setattr(module, "_resolve_repo_context", _context)
+ # Only a .py change exists; restricting to .md counts zero files.
+ result = await module.get_change_risk(extensions=["md"], baseline=0)
+
+ assert result["features"]["nf"] == 0
+ assert "warning" in result
+ assert "no counted file changes" in result["warning"].lower()
+
+
+async def test_get_change_risk_rejects_repo_all() -> None:
+ module = importlib.import_module("repowise.server.mcp_server.tool_change_risk")
+ result = await module.get_change_risk(repo="all")
+
+ assert "error" in result
+ assert "get_change_risk" in result["error"]