Repository navigation
Description
What is slow
Every time the branch picker in the thread toolbar opens, the client sends a local-only vcs.refreshStatus and then reloads the branch list with refresh: true, whatever the refresh found. That is two vcs.listRefs requests (6 git, about 190 ms of server work) and a cleared persisted branch cache for the whole environment, on every open, even when no ref has moved. #40385 already stopped this reload for the Git actions menu by comparing a refs digest. The picker uses its own command, refreshLocalStatusForPicker, which still invalidates refs unconditionally.
Measured
- Windows Server 2025, 4 vCPU, 16 GB, Defender real-time scanning on, Git 2.56.1.windows.1, commit 7d3677f.
- A 32,168-file clone with a local bare remote, on branch
workwith 5 local and 5 remote branches. 113 replay threads in two projects. Headless Chromium on the thread page. Nothing changed in the repository between opens. - 7 picker opens (open, wait 4 s, Escape, wait 3 s). 7 of 7 sent the same requests:
+58 vcs.refreshStatus {"cwd":"C:\\p\\repos\\p0","localOnly":true}
+1105 vcs.listRefs {"cwd":"C:\\p\\repos\\p0","query":"work","refresh":true,"limit":10}
+1105 vcs.listRefs {"cwd":"C:\\p\\repos\\p0","refresh":true,"limit":100}
- Git per open, from the server trace:
ws.rpc.vcs.refreshStatus10 git (886-1167 ms, median 1009 ms),ws.rpc.vcs.listRefs6 git (177-209 ms, median 196 ms). 16 git per open in 6 of 7 runs. In the other run a background poll landed in the window and the refresh ran 16. - The picker list showed in 124-361 ms (median 145 ms) from the cached list, so the reload only replaces a list that was already right.
How to measure it again
Harness (Windows, PowerShell; the next agent starts from a fresh machine)
- Check out the commit in
C:/p/wt. Runvp i, thenvp buildinapps/web. - Make two large repositories with a remote:
git clone --bare --no-local <a t3code checkout> C:/p/repos/remote.git, then for p0 and p1:git clone C:/p/repos/remote.git C:/p/repos/pN, set user.name/user.email,git checkout -b work; git push -u origin work, andgit branch feat-a HEAD~20; git push -u origin feat-a(same for feat-b, feat-c). Each has about 32,000 files. - Before the first server start, write
C:/p/home/userdata/settings.json:{"providerInstances":{"devReplay":{"displayName":"Replay (dev)","driver":"acpRegistry","config":{"source":"local","commandPath":"<output of node -p process.execPath>","commandArgs":["C:/p/wt/apps/server/scripts/dev-replay-agent.ts"]}}}} - In
apps/server, withT3CODE_DEV_SEED=0, runnode src/bin.ts serve --base-dir C:/p/home --port 4800 --host 127.0.0.1and keep it running (over SSH, start it withInvoke-CimMethod Win32_Process Create; it listens after about 50 s). node src/bin.ts project add C:/p/repos/p0 --base-dir C:/p/home(and p1).- Seed threads: from inside the repository,
node C:/p/wt/apps/server/src/bin.ts thread start --base-dir C:/p/home --provider devReplay --model command-output C:/p/repos/p0 goprints a thread idmcp:<uuid>. This run seeded 113 threads (73 in p0, 40 in p1). Wait until the server is idle before measuring. The page path is/ENVID/mcp:<uuid>; common.cjs fills inuserdata/environment-id. - Pair:
node src/bin.ts pair --ttl 1h --base-dir C:/p/homeprints a URL ending inpair#token=...; replace its host with 127.0.0.1:4800 and export it asPAIR. ExportPWas the checkout'snode_modules/.pnpm/playwright@*/node_modules/playwrightdirectory. - Delete
C:/p/home/userdata/logs/server.trace.ndjson*before a run; afterwards copy them to a directory and run sum.cjs over it. Each git process is arunGitCommandorVcsProcess.spawnspan.
common.cjs:
const fs = require('fs');
const { chromium } = require(process.env.PW);
const NL = String.fromCharCode(10);
exports.open = async (path) => {
const browser = await chromium.launch();
const page = await (await browser.newContext({ viewport: { width: 1600, height: 1000 } })).newPage();
const sent = [];
page.on('websocket', (ws) => ws.on('framesent', (f) => {
try { for (const line of String(f.payload).split(NL)) { if (!line) continue; const j = JSON.parse(line); if (j._tag === 'Request') sent.push({ t: Date.now(), tag: j.tag, p: JSON.stringify(j.payload).slice(0, 160) }); } } catch {}
}));
await page.goto(process.env.PAIR);
await page.waitForTimeout(4000);
const env = fs.readFileSync('C:/p/home/userdata/environment-id', 'utf8').trim();
await page.goto('http://127.0.0.1:4800' + path.replace('ENVID', env));
return { browser, page, sent };
};
exports.record = (tag, rec) => { fs.mkdirSync('C:/p/perf/out', { recursive: true }); fs.appendFileSync('C:/p/perf/out/' + tag + '.jsonl', JSON.stringify(rec) + NL); };
exports.reqs = (sent, a, b) => { const c = {}; for (const s of sent) if (s.t >= a && s.t <= b) c[s.tag] = (c[s.tag] || 0) + 1; return c; };
exports.seq = (sent, a, b) => sent.filter((s) => s.t >= a && s.t <= b && s.tag !== 'server.reportClientActivity').map((s) => '+' + (s.t - a) + ' ' + s.tag + ' ' + s.p);sum.cjs (node sum.cjs <traceDir> <out.jsonl>; CALLERS=1 lists the calling span):
const fs = require('fs'); const path = require('path');
const NL = String.fromCharCode(10);
const [dir, outFile, sf = 'start', ef = 'end'] = process.argv.slice(2);
const spans = [];
for (const f of fs.readdirSync(dir)) for (const l of fs.readFileSync(path.join(dir, f), 'utf8').split(NL)) { if (!l) continue; try { spans.push(JSON.parse(l)); } catch {} }
const byId = new Map(spans.map((s) => [s.spanId, s]));
const ms = (s) => Number(BigInt(s.startTimeUnixNano) / 1000000n);
const plumbing = ['VcsProcess', 'runGitCommand', 'GitVcsDriver.execute', 'GitVcsDriver.run', 'GitVcsDriver.gitCommand', 'gitCommand', 'ProcessRunner'];
const root = (s) => { while (s.parentSpanId && byId.has(s.parentSpanId)) s = byId.get(s.parentSpanId); return s; };
const caller = (s) => { let p = s; while (p.parentSpanId && byId.has(p.parentSpanId)) { p = byId.get(p.parentSpanId); if (!plumbing.some((x) => p.name.startsWith(x))) return p.name; } return '?'; };
const spawns = spans.filter((s) => s.name === 'VcsProcess.spawn' || s.name === 'runGitCommand');
for (const l of fs.readFileSync(outFile, 'utf8').split(NL)) {
if (!l) continue; const r = JSON.parse(l); const a = r[sf], b = r[ef];
const roots = new Map();
for (const sp of spawns) { const t = ms(sp); if (t < a || t > b) continue; const rt = root(sp); const e = roots.get(rt.name) || { n: 0, ms: new Set(), callers: {} }; e.n++; e.ms.add(Math.round(rt.durationMs)); const c = caller(sp); e.callers[c] = (e.callers[c] || 0) + 1; roots.set(rt.name, e); }
let total = 0; for (const e of roots.values()) total += e.n;
console.log('run', r.run, r.target || '', 'total git=' + total);
for (const [k, e] of roots) console.log(' ', k, 'git=' + e.n, 'ms=' + [...e.ms].join(','), process.env.CALLERS ? JSON.stringify(e.callers) : '');
}picker.cjs (node picker.cjs /ENVID/<thread id> with RUNS=7 SEQ=1 TAG=picker), then node sum.cjs <traces> C:/p/perf/out/picker.jsonl:
const { open, record, reqs, seq } = require('./common.cjs');
(async () => {
const runs = +(process.env.RUNS || 7); const tag = process.env.TAG;
const { browser, page, sent } = await open(process.argv[2]);
const trigger = page.getByRole('combobox', { name: /^Branch:/ }).first();
await trigger.waitFor({ timeout: 60000 });
await page.waitForTimeout(10000);
for (let i = 0; i < runs; i++) {
const t0 = Date.now();
await trigger.click();
await page.getByRole('option').first().waitFor({ timeout: 30000 });
const t1 = Date.now();
await page.waitForTimeout(4000);
await page.keyboard.press('Escape');
await page.waitForTimeout(3000);
const t2 = Date.now();
const r = { run: i, start: t0, end: t2, shown: t1 - t0, reqs: reqs(sent, t0, t2) };
if (process.env.SEQ) r.seq = seq(sent, t0, t2);
record(tag, r); console.log(JSON.stringify(r));
}
await browser.close();
})();Where the time goes
apps/web/src/components/BranchToolbarBranchSelector.tsx,handleOpenChange(line 683): on every open, it callsrefreshLocalStatusForPicker({ cwd, localOnly: true }).packages/client-runtime/src/state/vcs.ts,refreshLocalStatusForPicker(line 353):onSettled: invalidateRefs, which callsinvalidateCachedVcsRefs. That clears the persisted refs cache for the environment (cache.clearVcsRefs(environmentId)) and bumps the refs revision, so both mountedlistRefsqueries refetch withrefresh: true.refreshStatusabove it (line 339) usesinvalidateCachedVcsRefsIfChangedwith the server'srefsDigestinstead (perf(client-runtime): Git menu status refresh keeps the branch list when no ref moved #40385). The server returns no digest forlocalOnlyrefreshes:VcsStatusBroadcaster.refreshStatusskipsreadRefSnapshotwhenoptions.localOnlyis set.
Why it is avoidable
The branch list can only be stale if a ref moved. The refs-digest check that already runs for the Git menu answers that with one for-each-ref read. The picker open throws away a correct list and reads it again, every time.
What a fix has to show
The same 7 picker opens on an unchanged repository, on Windows:
- Expected: no
vcs.listRefsrequest on an open when no ref moved since the last refresh, so 10-11 git per open instead of 16 (6 fewer listRefs git; a digest read may add 1). The persisted branch cache survives the open. - What stays: the local-only status refresh itself (10 git, about 1 s here), which this issue does not question.
- Not regressed: a branch created, deleted or moved outside T3 Code (for example
git branch new-xin a terminal) appears in the list on the next picker open.
Found by the automated Windows performance sweep on 7d3677f.
Found by the automated Windows performance sweep on 7d3677f.
Activity