@kabmo-labs/kabmo 0.5.0 → 0.5.2

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kabmo-labs/kabmo",
-  "version": "0.5.0",
+  "version": "0.5.2",
   "description": "Web console and diagnostic workbench for AI coding agents — Elysia + Preact on Bun.",
   "license": "UNLICENSED",
   "author": "Wijanarko Putra Rajeb <wijanarko.rajeb@gmail.com>",

package/src/client/components/layout/session-sidebar/SessionItem.tsx

@@ -181,13 +181,10 @@ export function SessionItem({
               ) : status === 'stream' ? (
                 // The run spinner is the ONE thing on this row that says a run
                 // is in flight, and it was an unlabelled icon: a screen reader
-                // got a bare `svg`, and a test had no accessible way to ask
-                // "is this session running". `waiting-for.e2e.ts` needs exactly
-                // that as its gate — the wait picker only appears once this
-                // session's `stream` row has reached the client, so asserting
-                // the picker without asserting the row was a race measured on
-                // the Windows runner (the dependency's 20s run expired while the
-                // picker's own 20s timeout was still counting).
+                // got a bare `svg`, and nothing could ask "is this session
+                // running" by name. The rename and waiting cues beside it were
+                // already labelled, so this was the one status a reader — or an
+                // assertion — had to identify by shape.
                 <span role="status" aria-label="Running">
                   <Loader2 size={12} className="animate-spin" />
                 </span>

package/src/client/components/mobile/mobile-session-sidebar/SessionRow.tsx

@@ -159,7 +159,7 @@ export function MobileSessionRow({
               ) : status === 'stream' ? (
                 // Same label the desktop row carries, for the same reason: the
                 // phone draws the identical cue, and an unlabelled spinner is
-                // both unreachable to a screen reader and unassertable.
+                // unreachable to a screen reader.
                 <span role="status" aria-label="Running">
                   <Loader2 size={13} className="text-ink/50 animate-spin" />
                 </span>

package/src/server/lib/fs/git-actions-sync.test.ts

@@ -22,13 +22,30 @@
  * `MERGE_HEAD` file are the whole subject.
  */
 
-import { afterAll, describe, expect, test } from 'bun:test';
+import { afterAll, describe, expect, setDefaultTimeout, test } from 'bun:test';
 import fs from 'node:fs';
 import os from 'node:os';
 import path from 'node:path';
 import { GIT_BRANCH_ACTIONS } from '@/server/lib/fs/git-actions-branch';
 import { removeDir } from '@/test-support/remove-dir';
 
+/**
+ * The suite drives REAL git processes — `init`, `clone`, `commit`, `push`,
+ * `fetch`, `pull`, `merge` — and one test of it pays for most of them. Bun's
+ * default per-test budget is 5s, which a Linux runner clears and a Windows
+ * runner does not: measured on `windows-latest`, "reconciles a diverged branch"
+ * timed out at 5001ms while the same suite is green locally in 1.4s. Process
+ * spawn on that runner costs several times what it does on POSIX, and the
+ * log line above the failure (`killed 1 dangling process`) is the timeout
+ * firing mid-`git`.
+ *
+ * A budget, not a skip: the behaviour — a diverged branch reconciled by a merge
+ * and then pushed — is identical on both platforms, and skipping it there would
+ * report the Sync button as unverified on a platform the CLI ships to. The same
+ * budget is stated for the same reason in `git-update.server.test.ts`.
+ */
+setDefaultTimeout(30_000);
+
 const tempDirs: string[] = [];
 
 function tempDir(): string {