From e4d85bd0377b6455d05e6bd8ddbbd870f111e19d Mon Sep 17 00:00:00 2001
From: David Heinemeier Hansson <david@hey.com>
Date: Sun, 9 Aug 2026 13:19:46 +0200
Subject: [PATCH] Keep concurrent usage collectors off one shared temp file
 (#6654)

Two Claude collectors running at once both wrote the cache through a temp
path derived from the target, so the second replace found the file already
moved away and crashed the update with a FileNotFoundError.

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 bin/omarchy-agent-usage-claude                | 19 +++++--
 .../agent-usage-claude-scanner-test.sh        | 52 +++++++++++++++++++
 2 files changed, 68 insertions(+), 3 deletions(-)

diff --git a/bin/omarchy-agent-usage-claude b/bin/omarchy-agent-usage-claude
index 92535b07..c01bab39 100755
--- a/bin/omarchy-agent-usage-claude
+++ b/bin/omarchy-agent-usage-claude
@@ -23,6 +23,7 @@ import os
 import re
 import sqlite3
 import sys
+import tempfile
 import time
 import urllib.error
 import urllib.request
@@ -230,9 +231,21 @@ def read_fresh_json(path: Path, max_age_seconds: float) -> dict[str, Any] | None
 
 
 def write_json(path: Path, payload: dict[str, Any]) -> None:
-  tmp = path.with_suffix(path.suffix + ".tmp")
-  tmp.write_text(json.dumps(payload, separators=(",", ":"), sort_keys=True) + "\n", encoding="utf-8")
-  tmp.replace(path)
+  # A temp name unique to this writer, not derived from the target: several
+  # collectors can run at once (the update command backgrounds one per agent,
+  # the panel refreshes on its own), and a shared temp path means the second
+  # replace finds the first one's file already moved away.
+  handle_fd, tmp_name = tempfile.mkstemp(dir=path.parent, prefix=path.name + ".", suffix=".tmp")
+  tmp = Path(tmp_name)
+  try:
+    with os.fdopen(handle_fd, "w", encoding="utf-8") as handle:
+      handle.write(json.dumps(payload, separators=(",", ":"), sort_keys=True) + "\n")
+    # mkstemp opens at 0600; these caches were world-readable before.
+    tmp.chmod(0o644)
+    tmp.replace(path)
+  except BaseException:
+    tmp.unlink(missing_ok=True)
+    raise
 
 
 def cached_scan(projects_path: Path, max_age_seconds: float) -> dict[str, Any]:
diff --git a/test/shell.d/agent-usage-claude-scanner-test.sh b/test/shell.d/agent-usage-claude-scanner-test.sh
index 92da91eb..84beb807 100644
--- a/test/shell.d/agent-usage-claude-scanner-test.sh
+++ b/test/shell.d/agent-usage-claude-scanner-test.sh
@@ -125,3 +125,55 @@ result=$(HOME="$PI_HOME" XDG_CACHE_HOME="$PI_HOME/.cache" XDG_DATA_HOME="$PI_HOM
 [[ $(jq -c '.modelUsage' <<<"$result") == '{"claude-omp":{"cacheCreationInputTokens":1,"cacheReadInputTokens":4,"inputTokens":20,"outputTokens":5},"claude-pi":{"cacheCreationInputTokens":2,"cacheReadInputTokens":3,"inputTokens":10,"outputTokens":4}}' ]] ||
   fail "Claude collector filters pi and omp sessions to Anthropic providers" "$result"
 pass "Claude collector counts pi and omp subscription usage"
+
+# Collectors overlap in practice: the update command backgrounds one per agent
+# while the panel refreshes on its own. Two writers aiming at one cache file
+# must both land, not trip over a shared temp path.
+race_output=$(python3 - "$ROOT/bin/omarchy-agent-usage-claude" "$TEST_HOME/race.json" <<'PY'
+import importlib.util
+import json
+import sys
+import threading
+from importlib.machinery import SourceFileLoader
+from pathlib import Path
+
+# The collector has no .py suffix, so name its loader explicitly.
+spec = importlib.util.spec_from_loader("collector", SourceFileLoader("collector", sys.argv[1]))
+collector = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(collector)
+
+target = Path(sys.argv[2])
+failures = []
+start = threading.Barrier(8)
+
+def hammer(writer):
+  start.wait()
+  for round in range(25):
+    try:
+      collector.write_json(target, {"writer": writer, "round": round})
+    except Exception as error:
+      failures.append(repr(error))
+
+threads = [threading.Thread(target=hammer, args=(writer,)) for writer in range(8)]
+for thread in threads:
+  thread.start()
+for thread in threads:
+  thread.join()
+
+leftovers = sorted(path.name for path in target.parent.glob(target.name + ".*"))
+print(json.dumps({
+  "failures": failures[:3],
+  "mode": oct(target.stat().st_mode & 0o777),
+  "payload": json.loads(target.read_text(encoding="utf-8")),
+  "leftovers": leftovers,
+}))
+PY
+)
+
+[[ $(jq -c '.failures' <<<"$race_output") == "[]" ]] ||
+  fail "Claude collector survives concurrent writes to one cache file" "$race_output"
+[[ $(jq -r '.payload.writer != null and (.leftovers | length) == 0' <<<"$race_output") == "true" ]] ||
+  fail "Claude collector leaves one intact cache file and no temp files" "$race_output"
+[[ $(jq -r '.mode' <<<"$race_output") == "0o644" ]] ||
+  fail "Claude collector keeps cache files readable" "$race_output"
+pass "Claude collector survives concurrent writes to one cache file"