fix(engine): preserve Windows RSS samples after a query failure (#5357)

packages/engine/src/services/chromeMemorySampler.test.ts

@@ -1,4 +1,5 @@
 import { afterEach, describe, expect, it, vi } from "vitest";
+import { sampleProcessRss } from "../utils/processRss.js";
 import {
   createChromeMemorySampler,
   mergeSample,
@@ -170,6 +171,31 @@ describe("createChromeMemorySampler", () => {
     expect(sampler.stats()).toEqual(empty);
   });
 
+  it("reports Windows browser and GPU memory when the renderer query fails", async () => {
+    const exec = async (_file: string, args: readonly string[]) => {
+      const pid = Number(args[4]?.slice("PID eq ".length));
+      if (pid === 2) throw new Error("tasklist timed out");
+      return { stdout: `"chrome.exe","${pid}","Console","1","${pid * 10240} K"\r\n` };
+    };
+    const onSample = vi.fn();
+    const sampler = createChromeMemorySampler({
+      getPids: async () => ({ browser: 1, renderers: [2], gpu: [3] }),
+      sampleRss: (pids) => sampleProcessRss(pids, exec, "win32"),
+      intervalMs: 100,
+      onSample,
+    });
+
+    await sampler.sampleOnce();
+    expect(sampler.stats()).toEqual({
+      browserRssPeakMb: 10,
+      rendererRssPeakMb: undefined,
+      rssLastMb: 40,
+      gpuProcessSeenLastSample: true,
+      samples: 1,
+    });
+    expect(onSample).toHaveBeenCalledWith(sampler.stats());
+  });
+
   it("unrefs the interval so a forgotten stop() cannot hold the process open", () => {
     // Real timers: a real NodeJS.Timeout reports hasRef(); the interval is far
     // enough out that no tick fires during the test.

packages/engine/src/utils/processRss.test.ts

@@ -115,6 +115,21 @@ describe("sampleProcessRss", () => {
     expect(result).toEqual([{ pid: 9, rssMb: 2 }]);
   });
 
+  it.each([1, 2, 3])("keeps successful Windows samples when pid %i fails", async (failedPid) => {
+    const calls: number[] = [];
+    const exec = async (_file: string, args: readonly string[]) => {
+      const pid = Number(args[4]?.slice("PID eq ".length));
+      calls.push(pid);
+      if (pid === failedPid) throw new Error("tasklist timed out");
+      return { stdout: `"chrome.exe","${pid}","Console","1","${pid * 1024} K"\r\n` };
+    };
+
+    expect(await sampleProcessRss([1, 2, 3], exec, "win32")).toEqual(
+      [1, 2, 3].filter((pid) => pid !== failedPid).map((pid) => ({ pid, rssMb: pid })),
+    );
+    expect(calls).toEqual([1, 2, 3]);
+  });
+
   it("drops invalid pids and duplicates before calling ps", async () => {
     const calls: Array<readonly string[]> = [];
     const exec = async (_file: string, args: readonly string[]) => {

packages/engine/src/utils/processRss.ts

@@ -91,8 +91,10 @@ export async function sampleProcessRss(
     if (platform === "win32") {
       const results: ProcessRssSample[] = [];
       for (const pid of unique) {
-        const { stdout } = await exec("tasklist", ["/FO", "CSV", "/NH", "/FI", `PID eq ${pid}`]);
-        results.push(...parseTasklistCsv(pid, stdout));
+        try {
+          const { stdout } = await exec("tasklist", ["/FO", "CSV", "/NH", "/FI", `PID eq ${pid}`]);
+          results.push(...parseTasklistCsv(pid, stdout));
+        } catch {}
       }
       return results;
     }