fix(gitattributes): every tracked text file is LF on every checkout (CH-6)

.gitattributes

@@ -8,6 +8,22 @@
 # platform. Both halves are required: the attribute governs checkout, the writer
 # governs regeneration.
 
+# Catch-all, not a list: a file with no extension, or an extension added later, must not
+# be the one that checks out CRLF. The rules below document intent and pin the cases.
+* text=auto eol=lf
+
+# Binaries are named so git's heuristic is not trusted for them.
+*.png binary
+*.gif binary
+*.mp4 binary
+*.ico binary
+*.jpg binary
+*.jpeg binary
+*.woff binary
+*.woff2 binary
+*.pdf binary
+.coverage binary
+
 .git/hooks/** text eol=lf
 *.sh text eol=lf
 *.ps1 text eol=lf

tests/test_gitattributes.py

@@ -0,0 +1,64 @@
+"""A tracked text file is never checked out CRLF; a tracked binary is never treated as text."""
+
+import subprocess
+from pathlib import Path
+
+import pytest
+
+ROOT = Path(__file__).resolve().parent.parent
+LF_ENTRY = "i/lf"
+BINARY_ENTRY = "i/-text"
+BINARY_TEXT_ATTRS = {"unset", "unspecified"}
+
+
+def _git(*args: str, stdin: str | None = None) -> str:
+    return subprocess.run(
+        ["git", *args],
+        cwd=ROOT,
+        input=stdin,
+        capture_output=True,
+        text=True,
+        check=True,
+    ).stdout
+
+
+def _tracked_by_index_state() -> dict[str, list[str]]:
+    by_state: dict[str, list[str]] = {}
+    for line in _git("ls-files", "--eol", "-z").split("\0"):
+        if not line:
+            continue
+        info, path = line.split("\t", 1)
+        by_state.setdefault(info.split()[0], []).append(path)
+    return by_state
+
+
+def _attr(attr: str, paths: list[str]) -> dict[str, str]:
+    out = _git("check-attr", "-z", "--stdin", attr, stdin="\0".join(paths) + "\0")
+    fields = out.split("\0")
+    return {fields[i]: fields[i + 2] for i in range(0, len(fields) - 2, 3)}
+
+
+@pytest.fixture(scope="module")
+def tracked() -> dict[str, list[str]]:
+    if not (ROOT / ".git").exists():
+        pytest.skip("not a git checkout")
+    return _tracked_by_index_state()
+
+
+def test_every_tracked_lf_text_file_is_pinned_to_lf(tracked):
+    paths = tracked.get(LF_ENTRY, [])
+    assert paths
+    wrong = {p: v for p, v in _attr("eol", paths).items() if v != "lf"}
+    assert not wrong, f"tracked text files without eol=lf: {sorted(wrong)}"
+
+
+def test_every_tracked_binary_file_is_not_text(tracked):
+    paths = tracked.get(BINARY_ENTRY, [])
+    assert paths
+    wrong = {p: v for p, v in _attr("text", paths).items() if v not in BINARY_TEXT_ATTRS}
+    assert not wrong, f"tracked binaries git may treat as text: {sorted(wrong)}"
+
+
+def test_no_tracked_file_is_crlf_in_the_index(tracked):
+    bad = [p for state in ("i/crlf", "i/mixed") for p in tracked.get(state, [])]
+    assert not bad, f"CRLF in the index: {sorted(bad)}"