From ff6ed7c491ba285178a2fdc0975d619aefae7035 Mon Sep 17 00:00:00 2001
From: Teknium <127238744+teknium1@users.noreply.github.com>
Date: Sat, 1 Aug 2026 14:35:35 -0700
Subject: [PATCH] fix(update): detect EOL-only churn via numstat, not name-only
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

_normalize_managed_eol isolated line-ending churn from real edits by
diffing twice: all dirty files minus files still dirty under
--ignore-cr-at-eol. But 'git diff --name-only --ignore-cr-at-eol'
computes its file list from blob/stat differences BEFORE the CR filter
is applied, so it still lists CR-only files. On git 2.48.1 the two
name-only sets are therefore identical, _eol_only() is always empty, and
a managed Windows checkout gets pinned to core.autocrlf=false with the
whole CRLF tree left dirty — breaking the next 'git checkout' on update
(the exact failure this function exists to prevent).

Compute the real-edit set with 'git diff --numstat --ignore-cr-at-eol'
instead: numstat honors the CR filter (a CR-only file produces no
record), so eol-only files are correctly identified and cleared while
genuine edits are preserved. Pin core.quotepath=false so non-ASCII paths
parse. Verified at 1200 files: 1199 eol-only normalized, one real edit
preserved, autocrlf pinned only after the tree reads clean.

This was a pre-existing failure on main (test_update_eol_churn's
test_churn_across_more_files_than_fit_in_one_argv failed deterministically
on git 2.48.1), surfaced while landing unrelated file-tools PRs.
---
 hermes_cli/update_cmd.py | 30 +++++++++++++++++++++++++++++-
 1 file changed, 29 insertions(+), 1 deletion(-)

diff --git a/hermes_cli/update_cmd.py b/hermes_cli/update_cmd.py
index d5e1c445f4..95982c0f28 100644
--- a/hermes_cli/update_cmd.py
+++ b/hermes_cli/update_cmd.py
@@ -3474,8 +3474,36 @@ def _normalize_managed_eol(git_cmd, repo_root):
             return None
         return {p for p in out.stdout.split("\0") if p}
 
+    def _real_dirty():
+        # Files with a *content* change once CRLF differences are ignored.
+        # NOTE: ``diff --name-only --ignore-cr-at-eol`` still LISTS CR-only
+        # files (the name list is computed from blob/stat differences before
+        # the CR filter is applied), so it cannot be used to isolate real
+        # edits. ``--numstat`` does honor the filter: a CR-only file produces
+        # no numstat record, while a genuinely-edited file does. Parse the
+        # paths out of numstat instead.
+        out = subprocess.run(
+            probe + ["-c", "core.quotepath=false",
+                     "diff", "--numstat", "--ignore-cr-at-eol"],
+            cwd=repo_root,
+            capture_output=True,
+            text=True, encoding="utf-8", errors="replace",
+        )
+        if out.returncode != 0:
+            return None
+        paths = set()
+        for line in out.stdout.splitlines():
+            if not line.strip():
+                continue
+            # Format: "<added>\t<deleted>\t<path>". Rename detection is off in
+            # plain diff, so there is exactly one path field per record.
+            parts = line.split("\t", 2)
+            if len(parts) == 3 and parts[2]:
+                paths.add(parts[2])
+        return paths
+
     def _eol_only():
-        all_dirty, real_dirty = _dirty(), _dirty("--ignore-cr-at-eol")
+        all_dirty, real_dirty = _dirty(), _real_dirty()
         if all_dirty is None or real_dirty is None:
             return None
         return all_dirty - real_dirty