docs(quickstart): re-transcribe both boot samples against CLI 17.4.0

[hotlong]

Part of #141

Refresh of content/docs/quickstart.mdx for @objectstack/cli 17.4.0. Both paths were re-booted as independent pty captures. Each raw capture was diffed against its published block, checking presence and order mechanically. Every difference was classified against a 17.3.0 control.

What changed, and what the 17.3.0 control says about it

Surface	At 17.4.0	17.3.0 control
Both pins (:63, :201)	17.3.0 → 17.4.0	—
Path A block	unchanged. 0 published lines absent, 0 order violations, 3 independent boots	—
Path B block	only the declared machine-variable Build complete timing: 143ms → 97ms, re-transcribed from this capture	—
Both declarations	SQL-driver schema-sync notes no longer print (0 undelivered: lines across all 17.4.0 captures), so they are dropped	5 per boot, so new at 17.4.0
Both declarations	two SQL-driver error lines now print: a DATABASE_ERROR for a missing _objectstack_sequences table (1 on Path A, 2 on Path B) and an Insert operation failed on sys_oauth_resource with a stack trace. Both are named.	0 in 4 control boots, so new at 17.4.0
Path B declaration	no Boot diagnostics block on Path B (0 of 12 boots), so it is dropped. Path A still prints one.	1 warning in 4 of 4 boots, so new at 17.4.0
Path B declaration	"MCP connection trailer above the banner" replaced: the position is a race (below)	varies at 17.3.0 as well, so pre-existing

The MCP trailer has no fixed position

The trailer is printed by the dev parent process when it receives the objectstack:listening IPC message. The banner is printed by the serve child, which (in publishBoundPort) sends that message and then prints the banner. Two processes write to one terminal. Measured on pre-signal snapshots, so post-Ctrl+C output is excluded:

• 17.4.0: 10 of 12 boots print it after Press Ctrl+C to stop. 2 of 12 interleave it into the banner line by line; both were under CPU contention.
• 17.3.0 control: 3 of 4 after, 1 of 4 interleaved.
• The 17.3.0 round's single capture had it between the dev-admin lines and Config:. That is the reading "above the banner" was written from.

The new wording says it usually follows Press Ctrl+C to stop but can land inside the banner. That describes what was measured, without pinning a position that no version holds.

The control

Installing @objectstack/cli@17.3.0 today does not reproduce 17.3.0. Its caret ranges pull core/runtime/objectql/driver-* forward to 17.4.0. Pinning every @objectstack/* to 17.3.0 through overrides is not enough either: plugin-auth@17.3.0 asks for better-auth ^1.7.2, which now resolves to a release whose @better-auth/core/db no longer exports createLocalAccountIssuer. AuthPlugin then fails to load and only 25 plugins boot, against the published 30. (17.4.0 pins the better-auth family to exact 1.7.2, so readers of this page are unaffected.)

The control was therefore resolved as of before better-auth 1.7.3 shipped: npm install --before=2026-09-06T03:00:00Z. Read back, all 55 @objectstack/* copies are at 17.3.0 and @better-auth/core is at 1.7.2. The Path B control scaffold is the documented init -t app --install under npm_config_before, and its cli/spec/runtime/objectql read back at 17.3.0. As a validation, the control's Path A capture carries every line of the published 17.3.0 block, with 0 absent and 0 order violations.

Upstream status of the two new error lines

Both are already filed upstream and closed after 17.4.0 was published (2026-09-09T03:58Z):

• The sys_oauth_resource bootstrap row is re-inserted on every boot and its UNIQUE refusal is now logged at ERROR — an idempotent step reporting as a failure objectstack#17176 covers the sys_oauth_resource line. Its change, fix(plugin-auth): build one better-auth instance per boot, so the RFC 8707 resource row is seeded once objectstack#17525, merged 2026-09-10.
• A SELECT ... WHERE 1 = 0 column-existence probe against _objectstack_sequences is logged at ERROR on a normal boot, before the table exists objectstack#17175 covers the _objectstack_sequences probe. That card closed on 2026-09-11.

latest is still 17.4.0 for cli, objectql, plugin-auth and driver-sql, so the page declares what a reader installs today. The next trigger round should see both lines leave. Dedup was done over the REST issue lists plus a local grep, and nothing new was filed.

Re-measured, no drift

• Port: 0 rebind messages in 12 uncontended 17.4.0 boots, all binding 3000. With a real os start serving 3000 (HTTP 200), ↪ server bound to port 3001 (requested 3000) appears exactly once and the API binds 3001. Same at the 17.3.0 control. The sentence at :212-214 stands.
• Original item 3: 0 > **Warning blockquotes on 75cc0f0. The live control is 1 other > ** blockquote present at :108.
• Seeded dev credential and its dev-only caveat: byte-identical on the page, at 17.4.0 and at 17.3.0. Transcribed as is.

Gates on 75cc0f0 (the final commit; git status --porcelain = 0 lines)

• lock + pnpm turbo run build --force → VERDICT command-exit 0, Tasks: 1 successful, 1 total, Cached: 0 cached, 1 total
• lock + pnpm turbo run test --force → VERDICT command-exit 0, Tasks: 1 successful, 1 total, Cached: 0 cached, 1 total
• lock + pnpm turbo run type-check --force --continue → VERDICT command-exit 0, Tasks: 1 successful, 1 total
• check-locale-surface.mjs EXIT=0 → … neither llms consumer carries a numeric character reference or a malformed link target
• check-translations.mjs EXIT=0 → translations gate passed
• check-translation-ownership.mjs --actor hotlong --files changed.txt EXIT=0 → This PR touches 0 translation artifact(s) and 1 other file(s).
• check-translation-output.mjs --self-test EXIT=0, and --files changed-output.txt EXIT=0 → translation output gate passed (145 pre-existing finding(s) reported)
• also check-node-floor.mjs (+ --self-test), check-half-states.mjs --self-test, gen-zh-hant.mjs --check: all EXIT=0

The rendered en/docs/quickstart.html from that build shows both new identifiers as inline code, Build complete (97ms), and 0 schema-sync.

Scope: content/docs/quickstart.mdx only. reference/cli.mdx and the locale siblings are untouched.

---

Generated by Claude Code