[v0.3.0] ingestion — pin CPython source by commit SHA

.planning/agent-context/cpython-source-sha-pin.md

@@ -57,11 +57,22 @@
 ## 5. Decision log
 
 - Resolved SHAs (tag → 40-hex commit), one line each:
-  - 3.10 / v3.10.20 →
-  - 3.11 / v3.11.15 →
-  - 3.12 / v3.12.13 →
-  - 3.13 / v3.13.13 →
-  - 3.14 / v3.14.4 →
-- Where/how the verification aborts on mismatch:
+  - 3.10 / v3.10.20 → 842e987df856a5d4db37933c62a3456930a19092
+  - 3.11 / v3.11.15 → 2340a037f7450e70fccfe411e6531afb4d57a312
+  - 3.12 / v3.12.13 → 3bb231a6a5dc02b95658877318bf61501a7209e9
+  - 3.13 / v3.13.13 → 01104ce1beb3135c2e0c01ec835b994c1f55a1c0
+  - 3.14 / v3.14.4 → 23116f998f6789d8c2fbe5ed5b8146854c8c2a4f
+- Where/how the verification aborts on mismatch: after the shallow
+  tag-based clone in `build-index`, `git -C <clone_dir> rev-parse HEAD` is
+  compared to the authoritative config SHA. A mismatch logs the version, tag,
+  actual SHA, and expected SHA, then raises `SystemExit(1)` before Sphinx setup
+  or content ingestion can proceed.
 - **Draft SECURITY.md threat-model paragraph (for Vision to apply):**
-  >
+  > The largest build-time supply-chain input is the `build-index` clone of the
+  > upstream CPython repository, which provides the source tree used to generate
+  > canonical documentation content. Each supported CPython docs release is pinned
+  > to the exact commit SHA that its human-readable release tag currently resolves
+  > to; the tag is retained for operator readability, but the SHA is the
+  > authoritative integrity anchor. If a tag is reissued, moved, or otherwise
+  > resolves to different source content, the build fails before Sphinx setup or
+  > content ingestion rather than silently publishing changed documentation.

src/mcp_server_python_docs/__main__.py

@@ -65,6 +65,8 @@ def _consume_saved_stdout_fd() -> int:
 logger = logging.getLogger("mcp_server_python_docs")
 
 # === Now safe to import everything else ===
+import subprocess  # noqa: E402
+
 import click  # noqa: E402
 
 from mcp_server_python_docs.ingestion.cpython_versions import (  # noqa: E402
@@ -110,6 +112,33 @@ def serve() -> None:
         pass  # Client disconnected (HYGN-03)
 
 
+def _verify_cpython_source_sha(
+    clone_dir: str,
+    *,
+    version: str,
+    tag: str,
+    expected_sha: str,
+) -> None:
+    """Abort the docs build if a CPython tag resolves to unexpected content."""
+    rev_parse = subprocess.run(
+        ["git", "-C", clone_dir, "rev-parse", "HEAD"],
+        check=True,
+        capture_output=True,
+        text=True,
+    )
+    actual_sha = rev_parse.stdout.strip()
+    if actual_sha != expected_sha:
+        logger.error(
+            "CPython %s source integrity check failed: tag %s "
+            "resolved to %s, expected %s. Aborting build.",
+            version,
+            tag,
+            actual_sha,
+            expected_sha,
+        )
+        raise SystemExit(1)
+
+
 @main.command("build-index")
 @click.option(
     "--versions",
@@ -124,7 +153,6 @@ def serve() -> None:
 def build_index(versions: str, skip_content: bool) -> None:
     """Build the documentation index from objects.inv and Sphinx JSON."""
     import shutil
-    import subprocess
     import tempfile
     import venv
     from pathlib import Path
@@ -224,6 +252,12 @@ def build_index(versions: str, skip_content: bool) -> None:
                         capture_output=True,
                         text=True,
                     )
+                    _verify_cpython_source_sha(
+                        clone_dir,
+                        version=version,
+                        tag=config["tag"],
+                        expected_sha=config["sha"],
+                    )
 
                     # Create dedicated Sphinx venv (INGR-C-02)
                     venv_dir = os.path.join(clone_dir, "_sphinx_venv")

src/mcp_server_python_docs/ingestion/cpython_versions.py

@@ -8,6 +8,7 @@ class CPythonDocsBuildConfig(TypedDict):
     """Build settings for one CPython documentation release."""
 
     tag: str
+    sha: str
     sphinx_pin: str
 
 
@@ -21,12 +22,32 @@ class CPythonDocsBuildConfig(TypedDict):
 
 SUPPORTED_DOC_VERSIONS_CSV: Final[str] = ",".join(SUPPORTED_DOC_VERSIONS)
 
-# CPython git tags are pinned so content builds are reproducible and do not
-# drift when a maintenance branch receives new commits.
+# CPython git SHAs are authoritative for content build integrity. Tags are kept
+# for human-readable version mapping, but a moved tag must fail verification.
 CPYTHON_DOCS_BUILD_CONFIG: Final[dict[str, CPythonDocsBuildConfig]] = {
-    "3.10": {"tag": "v3.10.20", "sphinx_pin": "sphinx==3.4.3"},
-    "3.11": {"tag": "v3.11.15", "sphinx_pin": "sphinx~=7.2.0"},
-    "3.12": {"tag": "v3.12.13", "sphinx_pin": "sphinx~=8.2.0"},
-    "3.13": {"tag": "v3.13.13", "sphinx_pin": "sphinx<9.0.0"},
-    "3.14": {"tag": "v3.14.4", "sphinx_pin": "sphinx<9.0.0"},
+    "3.10": {
+        "tag": "v3.10.20",
+        "sha": "842e987df856a5d4db37933c62a3456930a19092",
+        "sphinx_pin": "sphinx==3.4.3",
+    },
+    "3.11": {
+        "tag": "v3.11.15",
+        "sha": "2340a037f7450e70fccfe411e6531afb4d57a312",
+        "sphinx_pin": "sphinx~=7.2.0",
+    },
+    "3.12": {
+        "tag": "v3.12.13",
+        "sha": "3bb231a6a5dc02b95658877318bf61501a7209e9",
+        "sphinx_pin": "sphinx~=8.2.0",
+    },
+    "3.13": {
+        "tag": "v3.13.13",
+        "sha": "01104ce1beb3135c2e0c01ec835b994c1f55a1c0",
+        "sphinx_pin": "sphinx<9.0.0",
+    },
+    "3.14": {
+        "tag": "v3.14.4",
+        "sha": "23116f998f6789d8c2fbe5ed5b8146854c8c2a4f",
+        "sphinx_pin": "sphinx<9.0.0",
+    },
 }

tests/test_ingestion.py

@@ -9,8 +9,10 @@
 
 import io
 import os
+import re
 import runpy
 import shutil
+import subprocess
 import sys
 import types
 
@@ -51,8 +53,45 @@ def test_supported_versions_have_pinned_docs_build_config(self):
         for version in SUPPORTED_DOC_VERSIONS:
             config = CPYTHON_DOCS_BUILD_CONFIG[version]
             assert config["tag"].startswith(f"v{version}.")
+            assert re.fullmatch(r"[0-9a-f]{40}", config["sha"])
             assert config["sphinx_pin"].startswith("sphinx")
 
+    def test_cpython_source_sha_verification_aborts_on_mismatch(
+        self,
+        monkeypatch,
+        caplog,
+    ):
+        from mcp_server_python_docs import __main__ as cli_main
+
+        calls: list[list[str]] = []
+
+        def fake_run(
+            cmd: list[str],
+            *,
+            check: bool,
+            capture_output: bool,
+            text: bool,
+        ) -> subprocess.CompletedProcess[str]:
+            calls.append(cmd)
+            assert check is True
+            assert capture_output is True
+            assert text is True
+            return subprocess.CompletedProcess(cmd, 0, stdout="b" * 40 + "\n")
+
+        monkeypatch.setattr(cli_main.subprocess, "run", fake_run)
+
+        with pytest.raises(SystemExit) as exc_info:
+            cli_main._verify_cpython_source_sha(
+                "/tmp/cpython-3.14",
+                version="3.14",
+                tag="v3.14.4",
+                expected_sha="a" * 40,
+            )
+
+        assert exc_info.value.code == 1
+        assert calls == [["git", "-C", "/tmp/cpython-3.14", "rev-parse", "HEAD"]]
+        assert "source integrity check failed" in caplog.text
+
 
 class TestJsonBuildRequirements:
     def test_omits_html_only_sphinx_extensions(self, tmp_path):