[BUG] git/gh subprocess text is locale-dependent; pin the C locale via iris.shell.git_env()

iris/analysis/code_provenance.py

@@ -23,6 +23,8 @@
 
 import re
 import subprocess
+
+from iris.shell import git_env
 from collections import defaultdict
 from dataclasses import dataclass
 from datetime import datetime, timedelta, timezone
@@ -198,6 +200,7 @@ def _blame_file_ages(
             capture_output=True,
             text=True,
             timeout=BLAME_TIMEOUT_SECONDS,
+            env=git_env(),
         )
         if result.returncode != 0:
             return None

iris/analysis/durability.py

@@ -21,6 +21,8 @@
 
 import re
 import subprocess
+
+from iris.shell import git_env
 from collections import defaultdict
 from dataclasses import dataclass, field
 from datetime import datetime, timezone
@@ -255,6 +257,7 @@ def _run_blame(repo_path: str, file_path: str) -> list[str] | None:
             capture_output=True,
             text=True,
             timeout=BLAME_TIMEOUT_SECONDS,
+            env=git_env(),
         )
         if result.returncode != 0:
             return None

iris/analysis/priming_detector.py

@@ -20,6 +20,8 @@
 
 import os
 import subprocess
+
+from iris.shell import git_env
 from dataclasses import dataclass
 from datetime import datetime
 
@@ -147,6 +149,7 @@ def _get_introduction_date(repo_path: str, file_path: str) -> datetime | None:
             capture_output=True,
             text=True,
             timeout=10,
+            env=git_env(),
         )
         if result.returncode != 0 or not result.stdout.strip():
             return None

iris/analysis/repo_kind.py

@@ -26,6 +26,8 @@
 
 import os
 import subprocess
+
+from iris.shell import git_env
 from enum import Enum
 
 
@@ -147,6 +149,7 @@ def _tracked_files(repo_path: str) -> list[str]:
             errors="surrogateescape",
             check=True,
             timeout=60,
+            env=git_env(),
         )
     except (subprocess.SubprocessError, OSError):
         return []

iris/cli.py

@@ -8,6 +8,7 @@
 from collections.abc import Callable
 from datetime import datetime, timedelta, timezone
 
+from iris.shell import git_env
 from iris.i18n import get_strings, SUPPORTED_LANGS
 from iris.platform.telemetry import span, record_metric, record_counter, record_duration, flush
 from iris.ingestion import window_cache
@@ -243,6 +244,7 @@ def _git_remote_url(repo_path: str) -> str | None:
         result = subprocess.run(
             ["git", "-C", repo_path, "remote", "get-url", "origin"],
             capture_output=True, text=True, check=True,
+            env=git_env(),
         )
         url = result.stdout.strip()
     except (subprocess.CalledProcessError, FileNotFoundError):
@@ -942,6 +944,7 @@ def _run_pr(argv: list[str]) -> None:
             result = subprocess.run(
                 ["gh", "pr", "view", "--json", "number", "-q", ".number"],
                 capture_output=True, text=True, check=True, cwd=repo_path,
+                env=git_env(),
             )
             pr_number = int(result.stdout.strip())
         except (subprocess.CalledProcessError, ValueError):
@@ -982,6 +985,7 @@ def _run_pr(argv: list[str]) -> None:
             subprocess.run(
                 ["gh", "pr", "comment", str(pr_number), "--body", markdown],
                 check=True, cwd=repo_path,
+                env=git_env(),
             )
             print(f"Comment posted on PR #{pr_number}.", file=sys.stderr)
         except subprocess.CalledProcessError:
@@ -1170,6 +1174,7 @@ def _hook_file_is_tracked(repo_path: str, hook_file: str) -> bool:
             cwd=repo_path,
             capture_output=True,
             timeout=10,
+            env=git_env(),
         )
     except (OSError, subprocess.SubprocessError):
         return False
@@ -1183,6 +1188,7 @@ def _hook_file_is_tracked(repo_path: str, hook_file: str) -> bool:
         capture_output=True,
         text=True,
         timeout=10,
+        env=git_env(),
     )
     if top.returncode != 0:
         return False

iris/hooks/manager.py

@@ -27,6 +27,7 @@
 import stat
 import subprocess
 import tempfile
+from iris.shell import git_env
 
 HOOK_NAME = "prepare-commit-msg"
 POST_COMMIT_HOOK_NAME = "post-commit"
@@ -130,6 +131,7 @@ def _git(repo_path: str, *args: str) -> str | None:
             capture_output=True,
             text=True,
             timeout=_PROBE_TIMEOUT_SECONDS,
+            env=git_env(),
         )
     except (OSError, subprocess.SubprocessError):
         return None

iris/ingestion/diff_reader.py

@@ -13,6 +13,8 @@
 
 import re
 import subprocess
+
+from iris.shell import git_env
 from dataclasses import dataclass
 
 from iris.models.commit import Commit
@@ -80,6 +82,7 @@ def read_commit_diff(repo_path: str, commit_hash: str) -> CommitDiff | None:
             capture_output=True,
             text=True,
             timeout=DIFF_TIMEOUT_SECONDS,
+            env=git_env(),
         )
         if result.returncode != 0:
             return None

iris/ingestion/git_reader.py

@@ -11,6 +11,8 @@
 
 import re
 import subprocess
+
+from iris.shell import git_env
 from datetime import datetime, timedelta, timezone
 
 from iris.models.commit import Commit, FileChange
@@ -78,6 +80,7 @@ def read_commits(
             capture_output=True,
             text=True,
             check=True,
+            env=git_env(),
         )
     except subprocess.CalledProcessError as exc:
         raise RuntimeError(f"git log failed for {repo_path}: {exc}\n{exc.stderr}") from exc
@@ -99,6 +102,7 @@ def _has_commits(repo_path: str) -> bool:
         ["git", "-C", repo_path, "rev-parse", "--verify", "--quiet", "HEAD"],
         capture_output=True,
         text=True,
+        env=git_env(),
     )
     if result.returncode == 0:
         return True
@@ -140,6 +144,7 @@ def read_pr_commits(
             capture_output=True,
             text=True,
             check=True,
+            env=git_env(),
         )
     except subprocess.CalledProcessError:
         return []

iris/ingestion/github_reader.py

@@ -18,6 +18,8 @@
 import re
 import shutil
 import subprocess
+
+from iris.shell import git_env
 from datetime import datetime, timedelta, timezone
 
 from iris.ingestion import window_cache
@@ -39,6 +41,7 @@ def detect_github_remote(repo_path: str) -> str | None:
             capture_output=True,
             text=True,
             check=True,
+            env=git_env(),
         )
     except (subprocess.CalledProcessError, FileNotFoundError):
         return None
@@ -231,6 +234,7 @@ def _fetch_pr_enrichment_graphql(
         try:
             result = subprocess.run(
                 args, capture_output=True, text=True, check=True,
+                env=git_env(),
             )
         except (subprocess.CalledProcessError, FileNotFoundError):
             return by_pr
@@ -296,6 +300,7 @@ def _gh_pr_list(nwo: str, fields: str, limit: int, gh_state: str) -> list[dict]
             capture_output=True,
             text=True,
             check=True,
+            env=git_env(),
         )
     except (subprocess.CalledProcessError, FileNotFoundError):
         return None
@@ -396,6 +401,7 @@ def read_single_pr(repo_path: str, pr_number: int) -> PullRequest | None:
             capture_output=True,
             text=True,
             check=True,
+            env=git_env(),
         )
     except (subprocess.CalledProcessError, FileNotFoundError):
         return None

iris/platform/config.py

@@ -2,6 +2,7 @@
 
 import json
 import os
+from iris.shell import git_env
 
 CONFIG_DIR = os.path.expanduser("~/.iris")
 CONFIG_FILE = os.path.join(CONFIG_DIR, "config.json")
@@ -66,6 +67,7 @@ def get_github_user() -> str | None:
         result = subprocess.run(
             ["gh", "api", "user", "-q", ".login"],
             capture_output=True, text=True, timeout=5,
+            env=git_env(),
         )
         if result.returncode == 0 and result.stdout.strip():
             user = result.stdout.strip()
@@ -78,6 +80,7 @@ def get_github_user() -> str | None:
             result = subprocess.run(
                 ["git", "config", "user.email"],
                 capture_output=True, text=True, timeout=5,
+                env=git_env(),
             )
             if result.returncode == 0 and result.stdout.strip():
                 user = result.stdout.strip()

iris/platform/identity.py

@@ -14,6 +14,8 @@
 
 import re
 import subprocess
+
+from iris.shell import git_env
 from datetime import datetime, timedelta, timezone
 
 from iris.models.commit import Commit
@@ -33,6 +35,7 @@ def _resolve_gh_name(username: str) -> str | None:
         r = subprocess.run(
             ["gh", "api", f"users/{username}", "-q", ".name"],
             capture_output=True, text=True, timeout=5,
+            env=git_env(),
         )
         if r.returncode == 0 and r.stdout.strip():
             return r.stdout.strip()
@@ -66,6 +69,7 @@ def _resolve_authors_via_commit_list(nwo: str, days: int) -> dict[str, str]:
                       '"\\(.commit.author.email)\\t\\(.author.login)"',
             ],
             capture_output=True, text=True, timeout=30,
+            env=git_env(),
         )
     except (FileNotFoundError, subprocess.TimeoutExpired):
         return {}
@@ -90,6 +94,7 @@ def _resolve_emails_via_repo(nwo: str, emails: set[str]) -> dict[str, str]:
                 ["gh", "api", f"repos/{nwo}/commits?author={email}&per_page=1",
                  "-q", ".[0].author.login"],
                 capture_output=True, text=True, timeout=10,
+                env=git_env(),
             )
             if r.returncode == 0 and r.stdout.strip():
                 result[email] = r.stdout.strip()

iris/shell.py

@@ -0,0 +1,26 @@
+"""Environment for the `git` and `gh` subprocesses the engine shells out to.
+
+Git localizes its messages through `LANG`/`LC_MESSAGES`. On a `pt_BR.UTF-8`
+machine, `git -C /missing log` fails with "Arquivo ou diretório inexistente";
+on the CI runner it fails with "No such file or directory". Everything the
+engine surfaces or matches from git's text — error messages in `RuntimeError`,
+tests asserting on them — therefore depended on the machine's locale.
+
+`git_env()` pins the C locale for those subprocesses only, without touching
+the user's shell. Structural output (`--porcelain`, `--numstat`, `--pretty`
+with separators, `--json`) was already locale-independent; this makes the
+human-readable remainder deterministic too.
+
+It is a function, not a module-level constant, so it reflects `os.environ`
+at call time (tests monkeypatch it) instead of a snapshot taken at import.
+"""
+
+import os
+
+
+def git_env() -> dict[str, str]:
+    """Return the current environment with the locale pinned to C."""
+    env = dict(os.environ)
+    env["LC_ALL"] = "C"
+    env["LANG"] = "C"
+    return env

tests/test_subprocess_locale.py

@@ -0,0 +1,73 @@
+"""Locale-independence of the git/gh subprocesses.
+
+Two guards: the runtime one proves git's error text reaches the caller in
+English even when the process runs under a Portuguese locale; the static one
+keeps every future `git`/`gh` `subprocess.run` on the pinned environment.
+
+Runnable as a plain script: `python tests/test_subprocess_locale.py`.
+"""
+
+import os
+import re
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from iris.ingestion.git_reader import read_commits
+from iris.shell import git_env
+
+_ROOT = Path(__file__).resolve().parent.parent / "iris"
+
+# A `subprocess.run(` whose argv literal starts with "git" or "gh". User-supplied
+# commands (external_reader) and non-git tools (curl in the self-updater) are
+# deliberately outside this rule.
+_GIT_CALL = re.compile(
+    r"subprocess\.run\((?P<body>.*?)\n\s*\)", re.DOTALL,
+)
+_STARTS_WITH_GIT = re.compile(r"""\[\s*["'](?:git|gh)["']""")
+
+
+def test_git_env_pins_c_locale_and_keeps_the_rest(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("LANG", "pt_BR.UTF-8")
+    monkeypatch.setenv("LC_ALL", "pt_BR.UTF-8")
+    monkeypatch.setenv("IRIS_PROBE", "kept")
+    env = git_env()
+    assert env["LC_ALL"] == "C"
+    assert env["LANG"] == "C"
+    assert env["IRIS_PROBE"] == "kept"
+    # Never mutates the caller's environment.
+    assert os.environ["LC_ALL"] == "pt_BR.UTF-8"
+
+
+def test_git_error_text_is_english_under_portuguese_locale(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    # Regression for the case that failed on pt_BR machines while passing in
+    # CI: the message git prints for a missing path.
+    monkeypatch.setenv("LANG", "pt_BR.UTF-8")
+    monkeypatch.setenv("LC_ALL", "pt_BR.UTF-8")
+    monkeypatch.setenv("LC_MESSAGES", "pt_BR.UTF-8")
+    with pytest.raises(RuntimeError) as exc:
+        read_commits(str(tmp_path / "does-not-exist"), days=90)
+    assert "no such file or directory" in str(exc.value).lower()
+
+
+def test_every_git_or_gh_subprocess_passes_the_pinned_env() -> None:
+    offenders: list[str] = []
+    for path in sorted(_ROOT.rglob("*.py")):
+        text = path.read_text()
+        for m in _GIT_CALL.finditer(text):
+            body = m.group("body")
+            if not _STARTS_WITH_GIT.search(body):
+                continue
+            if "env=" not in body:
+                line = text.count("\n", 0, m.start()) + 1
+                offenders.append(f"{path.relative_to(_ROOT.parent)}:{line}")
+    assert not offenders, "git/gh subprocess.run without env=git_env(): " + ", ".join(offenders)
+
+
+if __name__ == "__main__":
+    sys.exit(pytest.main([__file__, "-q"]))