fix(cli): stabilize Windows CI tests and rebalance slow shards (#12723)

* fix(cli): stabilize Windows CI tests and rebalance slow shards

Three Windows-only instabilities in the CLI unit suite:

1. httpapi-instance-route-auth.test.ts failed with an uncaught
   "Invalid handle" error. The test's ConfigProvider.layer(
   fromUnknown(...)) replaced the ambient config provider, blinding
   KILO_EXPERIMENTAL_DISABLE_FILEWATCHER=true that CI/preload sets. With
   the flag hidden, the @parcel/watcher Windows backend subscribed on the
   temp repo's .git; the tmpdir fixture then deleted that directory while
   the never-disposed per-test runtime still held the subscription, and
   CreateFileW failed with the hardcoded "Invalid handle" (napi rejection
   with no JS stack). Add the disable-filewatcher flag to every test
   config map that boots instances via the HttpApi app (instance-route-auth,
   cors, ui, exercise backend, kilo-edit, memory).

2. config-overlay.test.ts intermittently returned HTTP 500 on Windows.
   Filesystem.write's atomic temp-file+rename had no retry for Windows
   transient locked-file errors (EPERM/EACCES/EBUSY) from Defender/indexer
   and the detached background plugin install racing the rename in the same
   tmpdir. Mirror the proven cleanup.ts locked-error retry pattern with a
   short backoff, Windows-only.

3. Windows shards were badly imbalanced: the sharder weighted files by
   byte size, which concentrated every slow spawn/FS/lock-heavy file
   (snapshot, prompt, provider, run-process, instance-bootstrap,
   httpapi-session) into one shard (~612s vs ~356s siblings), and the
   resulting contention forced whole-file retries that doubled cost. Add
   TestShard.timedWeight and a committed test-timings.json seeded from CI
   junit data so shards balance by measured runtime (spread collapses from
   ~200s to ~18s) and contention-prone files spread across shards.
   Platforms without manifest entries fall back to size weighting.

* fix(cli): skip stale manifest entries in timed shard weighting

Bun.file().size returns 0 (never throws) for missing paths, so the
try/catch in timedWeight was dead code and stale/renamed manifest entries
added their time to the scale numerator with zero size, inflating the
size-to-time ratio that estimates unknown files. Skip entries with a
non-positive on-disk size instead of catching a throw that never happens.

* revert(cli): drop hardcoded test-timings manifest

The committed test-timings.json (482 entries) was a maintenance burden:
it goes stale as tests are added/renamed and no size-based heuristic can
replace it (slow subprocess outliers like run-process.test.ts are 7kb but
112s, 10x the runtime-per-byte of other files). Revert the timing-weighted
sharding to the prior size-based LPT. The Windows reliability fixes
(ConfigProvider filewatcher flag + Filesystem.write locked-file retry)
remain and are what eliminate the failures and the ~360s of retry overhead
that dominated the 12m50s shard. A maintainable runtime-based rebalance
(self-updating CI cache fed from the junit artifacts CI already uploads)
is a separate follow-up.

.changeset/windows-locked-file-retry.md

@@ -0,0 +1,5 @@
+---
+"@kilocode/cli": patch
+---
+
+Retry transient locked-file errors (EPERM/EACCES/EBUSY) on Windows when atomically saving config and other files. Background plugin installs and Windows Defender/indexer can briefly hold the temp file during the rename step, which previously surfaced as a 500 error. A short backoff now retries the rename so config writes succeed without surfacing the contention.

packages/opencode/src/util/filesystem.ts

@@ -66,6 +66,21 @@ function isEnoent(e: unknown): e is { code: "ENOENT" } {
   return typeof e === "object" && e !== null && "code" in e && (e as { code: string }).code === "ENOENT"
 }
 
+// kilocode_change start - Windows transient locked-file errors on atomic rename
+// Defender/indexer and concurrent writers (e.g. background plugin install) can
+// briefly hold the temp file, making MoveFileEx fail with EPERM/EACCES/EBUSY.
+// Retry with a short backoff instead of surfacing a 500; POSIX renames are atomic
+// so the retry path only fires under contention and never changes success semantics.
+function isLocked(e: unknown): boolean {
+  return (
+    typeof e === "object" &&
+    e !== null &&
+    "code" in e &&
+    ["EBUSY", "EACCES", "EPERM"].includes(String((e as { code: string }).code))
+  )
+}
+// kilocode_change end
+
 export async function write(p: string, content: string | Buffer | Uint8Array, mode?: number): Promise<void> {
   // kilocode_change start - atomic write via temp-file + rename to avoid partial reads on concurrent saves
   // Include a random suffix so that concurrent writes to the same path never share a temp file,
@@ -79,15 +94,23 @@ export async function write(p: string, content: string | Buffer | Uint8Array, mo
     }
     await rename(tmp, p)
   }
-  try {
-    await doWrite()
-  } catch (e) {
-    if (isEnoent(e)) {
-      await mkdir(dirname(p), { recursive: true })
+  const attempts = process.platform === "win32" ? 8 : 1
+  for (let attempt = 1; ; attempt++) {
+    try {
       await doWrite()
       return
+    } catch (e) {
+      if (isEnoent(e)) {
+        await mkdir(dirname(p), { recursive: true })
+        await doWrite()
+        return
+      }
+      if (isLocked(e) && attempt < attempts) {
+        await Bun.sleep(50 * attempt)
+        continue
+      }
+      throw e
     }
-    throw e
   }
   // kilocode_change end
 }

packages/opencode/test/kilocode/server/httpapi-kilo-edit.test.ts

@@ -30,7 +30,16 @@ const edit = {
 
 function app() {
   const handler = HttpRouter.toWebHandler(
-    HttpApiServer.routes.pipe(Layer.provide(ConfigProvider.layer(ConfigProvider.fromUnknown({})))),
+    // kilocode_change - keep the filewatcher-disable flag visible (see httpapi-instance-route-auth.test.ts)
+    HttpApiServer.routes.pipe(
+      Layer.provide(
+        ConfigProvider.layer(
+          ConfigProvider.fromUnknown({
+            KILO_EXPERIMENTAL_DISABLE_FILEWATCHER: process.env.KILO_EXPERIMENTAL_DISABLE_FILEWATCHER ?? "true",
+          }),
+        ),
+      ),
+    ),
     { disableLogger: true },
   ).handler
 

packages/opencode/test/kilocode/server/httpapi-memory.test.ts

@@ -14,7 +14,16 @@ type Json = Record<string, unknown>
 
 function app() {
   const handler = HttpRouter.toWebHandler(
-    HttpApiServer.routes.pipe(Layer.provide(ConfigProvider.layer(ConfigProvider.fromUnknown({})))),
+    // kilocode_change - keep the filewatcher-disable flag visible (see httpapi-instance-route-auth.test.ts)
+    HttpApiServer.routes.pipe(
+      Layer.provide(
+        ConfigProvider.layer(
+          ConfigProvider.fromUnknown({
+            KILO_EXPERIMENTAL_DISABLE_FILEWATCHER: process.env.KILO_EXPERIMENTAL_DISABLE_FILEWATCHER ?? "true",
+          }),
+        ),
+      ),
+    ),
     { disableLogger: true },
   ).handler
 

packages/opencode/test/server/httpapi-cors.test.ts

@@ -64,7 +64,16 @@ describe("HttpApi CORS", () => {
     Effect.gen(function* () {
       const handler = HttpRouter.toWebHandler(
         HttpApiApp.createRoutes().pipe(
-          Layer.provide(ConfigProvider.layer(ConfigProvider.fromUnknown({ KILO_SERVER_PASSWORD: "secret" }))),
+          // kilocode_change start - keep the filewatcher-disable flag visible (see httpapi-instance-route-auth.test.ts)
+          Layer.provide(
+            ConfigProvider.layer(
+              ConfigProvider.fromUnknown({
+                KILO_SERVER_PASSWORD: "secret",
+                KILO_EXPERIMENTAL_DISABLE_FILEWATCHER: process.env.KILO_EXPERIMENTAL_DISABLE_FILEWATCHER ?? "true",
+              }),
+            ),
+          ),
+          // kilocode_change end
         ),
         { disableLogger: true },
       ).handler

packages/opencode/test/server/httpapi-exercise/backend.ts

@@ -59,9 +59,15 @@ function app(modules: Runtime, options: CallOptions) {
   const web = HttpRouter.toWebHandler(
     modules.HttpApiApp.routes.pipe(
       Layer.provide(
+        // kilocode_change start - keep the filewatcher-disable flag visible (see httpapi-instance-route-auth.test.ts)
         ConfigProvider.layer(
-          ConfigProvider.fromUnknown({ KILO_SERVER_PASSWORD: password, KILO_SERVER_USERNAME: username }),
+          ConfigProvider.fromUnknown({
+            KILO_SERVER_PASSWORD: password,
+            KILO_SERVER_USERNAME: username,
+            KILO_EXPERIMENTAL_DISABLE_FILEWATCHER: process.env.KILO_EXPERIMENTAL_DISABLE_FILEWATCHER ?? "true",
+          }),
         ),
+        // kilocode_change end
       ),
     ),
     { disableLogger: true, memoMap: modules.memoMap },

packages/opencode/test/server/httpapi-instance-route-auth.test.ts

@@ -14,10 +14,15 @@ function app(input: { password?: string; username?: string }) {
     HttpApiApp.routes.pipe(
       Layer.provide(
         ConfigProvider.layer(
+          // kilocode_change start - keep the filewatcher-disable flag visible so the
+          // @parcel/watcher Windows backend does not subscribe on temp dirs that
+          // the tmpdir fixture deletes mid-test (throws "Invalid handle").
           ConfigProvider.fromUnknown({
             KILO_SERVER_PASSWORD: input.password,
             KILO_SERVER_USERNAME: input.username,
+            KILO_EXPERIMENTAL_DISABLE_FILEWATCHER: process.env.KILO_EXPERIMENTAL_DISABLE_FILEWATCHER ?? "true",
           }),
+          // kilocode_change end
         ),
       ),
     ),

packages/opencode/test/server/httpapi-ui.test.ts

@@ -52,12 +52,15 @@ function app(input?: { password?: string; username?: string }) {
   const handler = HttpRouter.toWebHandler(
     HttpApiApp.routes.pipe(
       Layer.provide(
+        // kilocode_change start - keep the filewatcher-disable flag visible (see httpapi-instance-route-auth.test.ts)
         ConfigProvider.layer(
           ConfigProvider.fromUnknown({
             KILO_SERVER_PASSWORD: input?.password,
             KILO_SERVER_USERNAME: input?.username,
+            KILO_EXPERIMENTAL_DISABLE_FILEWATCHER: process.env.KILO_EXPERIMENTAL_DISABLE_FILEWATCHER ?? "true",
           }),
         ),
+        // kilocode_change end
       ),
     ),
     { disableLogger: true },
@@ -100,12 +103,15 @@ function uiApp(input?: {
         input?.client ?? httpClient(new Response("ui")),
         RuntimeFlags.layer({ disableEmbeddedWebUi: input?.disableEmbeddedWebUi ?? false }),
         HttpServer.layerServices,
+        // kilocode_change start - keep the filewatcher-disable flag visible (see httpapi-instance-route-auth.test.ts)
         ConfigProvider.layer(
           ConfigProvider.fromUnknown({
             KILO_SERVER_PASSWORD: input?.password,
             KILO_SERVER_USERNAME: input?.username,
+            KILO_EXPERIMENTAL_DISABLE_FILEWATCHER: process.env.KILO_EXPERIMENTAL_DISABLE_FILEWATCHER ?? "true",
           }),
         ),
+        // kilocode_change end
       ]),
     ),
     { disableLogger: true },