playwright: per-request HTTP timeout so a wedged host fails fast (E1)

[borisbat]

Why

E1 from the integration-test sync investigation. On #118's windows-latest run, test_imgui_synth_ctrl_shortcuts wedged for 159s and tripped:

imgui_playwright.das:1404: daslang-live timed out (>120s)

i.e. it ran all the way to the popen watchdog (DEFAULT_TEST_TIMEOUT_SEC = 120) instead of failing fast.

Mechanism: the live-API client used dasHV's convenience GET/POST, which inherit libhv's default (effectively unbounded) response timeout. The host serves the live API on the GLFW main thread that also advances frames (see the header note in imgui_transport.das). So when a synth command wedges that thread, the host stops answering and stops rendering, and the client blocks on the POST with no timeout — the body can't make progress and the whole subprocess runs to the 120s watchdog. A mute 2-minute hang with no useful diagnostic.

What

Add http_get / http_post helpers in imgui_transport built on dasHV's existing request(req) rail (which exposes HttpRequest.timeout), with a hard per-request timeout (DEFAULT_REQUEST_TIMEOUT_SEC = 10s), and route every client call through them:

• live_api_transport's POST /command (every command + snapshot poll)
• wait_until_ready's GET /status
• get_text (→ reload / reset / read_status)
• post_signal (POST /shutdown / /reload / /reset)

Now a wedged host yields a null / status==0 reply in ~10s, so the wait_until budgets actually bound the failure and the diagnostic is legible instead of a silent 120s hang.

Scope / non-goals

• Client-side only — no host or daslang change. Uses dasHV's existing request(req) rail, so no dasHV add was needed; the convenience GET/POST stay as-is.
• This makes wedges fail fast and legibly; it does not fix the cause of the windows synth wedge (that's E2) or the test_docking_basic crash (0xC0000409, deferred). Both become much easier to diagnose once the harness fails in seconds rather than minutes.

Validation

• compile_check (through consuming tests) ✓, lint 0 issues ✓, das-fmt verify ✓.
• Functional validation is the ci: fix Windows caching + re-enable windows-latest #118 rebase on CI — windows-latest headless is the env where the wedge reproduces. The local harness needs a Release daslang-live this dev box doesn't currently have, so a local end-to-end smoke wasn't possible; the change is functionally equivalent to the old path plus a timeout (both go through libhv requests::request).

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR updates the Playwright/live-API client transport to apply a hard per-request HTTP timeout, so integration tests fail fast (≈10s) when the host wedges and stops responding, instead of hanging until the overall popen watchdog (120s).

Changes:

• Add http_get / http_post helpers in imgui_transport that set HttpRequest.timeout per request.
• Route live API calls in imgui_playwright (readiness polling, text fetches, and control signals) through these timeout-bounded helpers.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
widgets/imgui_transport.das	Introduces HttpReply, http_get, and http_post wrappers that set per-request timeouts and are used by live_api_transport.
widgets/imgui_playwright.das	Switches readiness polling, simple GETs, and signal POSTs to use the new timeout-bounded transport helpers.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.