ci(nests): defer cross-stack interop CI; document manual run

Removes the `hang-interop` + `browser-interop` jobs from
`build.yml` (added in commit `21947bc5` after a 10/10 stability
sweep × 22 tests = 220/220 hard-pass). Cold-cache cost is
~10 min hang + ~13 min browser; warm-cache critical-path add
is ~6 min via the parallel browser job. Most PRs don't touch
audio / MoQ / QUIC, so paying that cost on every PR is
net-negative for the change-pattern the repo sees.

Adds `nestsClient/tests/README.md` covering:
- when to run (changes to nip53 / moq-lite session / audio
  pipeline / MoqLiteNests* / ReconnectingNests* / :quic /
  the sidecars themselves);
- quick-start gradle commands for hang-only, browser-only,
  combined;
- prerequisites + first-run cache costs;
- all the configuration knobs (incl. the
  `-DnestsHangInteropTraceRelay=true` trace-capture switch
  added during the routing investigation);
- known limitations (hot-swap browser soft-pass,
  framesPerGroup pin-vs-prod gap, I7 cycle 2 truncation);
- a 4-step debug recipe for triaging a flaking scenario.

Plan updates:
- `2026-05-07-t16-closure-roadmap.md` — Priority 3 marked
  ⏸ DEFERRED instead of  CLOSED, pointing at the new README.
- `2026-05-07-cross-stack-interop-ci-gating.md` — status
  changed to ⏸ DEFERRED; YAML shape preserved verbatim in the
  plan for the next revisit.
- `2026-05-06-cross-stack-interop-test-results.md`,
  `2026-05-06-cross-stack-interop-test-gap-matrix.md` — CI
  integration § rewritten to "manual-run only" + README link.

The trace-capture instrumentation in `NativeMoqRelayHarness`
stays in place; it's useful for future flake triage even
without CI.
This commit is contained in:
Claude
2026-05-07 23:18:31 +00:00
parent 2d56b43672
commit 15b8560547
6 changed files with 246 additions and 201 deletions
@@ -92,9 +92,13 @@ DoD #5 (gap matrix coverage) closed.
tier — Chromium's `@moq/lite` 0.2.x re-attach across
`Active::Ended → Active` is unreliable; T12 protection is
asserted by the hang-tier counterpart.
- CI gating wired in
`2026-05-07-cross-stack-interop-ci-gating.md` after a 10/10
sweep × 22 tests = 220/220 pass stability bar.
- CI gating reached green-on-the-rig (10/10 × 22 = 220/220)
in commit `21947bc5`, then was deferred on cost grounds —
both suites run manually now per
[`nestsClient/tests/README.md`](../tests/README.md). YAML
shape preserved in
`2026-05-07-cross-stack-interop-ci-gating.md` for the next
revisit.
## Files referenced
@@ -440,25 +440,30 @@ relay-side timing under load.
## CI integration
**Wired** as of 2026-05-07 (commit `21947bc5`). Both
`hang-interop` and `browser-interop` jobs in
`.github/workflows/build.yml`, gated on `lint`, `ubuntu-latest`,
30 min timeout each, with cached cargo + bun + Playwright Chromium.
**Manual-run only** (deferred from CI on cost grounds). Both
suites are kept green and locally invokable; developer-facing
docs at [`nestsClient/tests/README.md`](../tests/README.md)
cover when/how/prerequisites.
Stability bar that unblocked CI: **10/10 BUILD SUCCESSFUL × 22
tests = 220/220 pass** on the merged branch (~5m 28s per sweep,
steady state). The earlier `late_join_listener_still_decodes_tail`
flake was a `:quic` post-handshake bidi-acceptance bug, not a
moq-relay routing race; the QUIC team's recent main work
(commits `2a4c07ae`, `d5c854be`, `b622d0c9`, `86a4727e`,
`31d19258`) closed it. See
`2026-05-07-moq-relay-routing-investigation.md` § Closure for
the trace evidence and
`2026-05-07-t16-closure-roadmap.md` for the rolled-up status.
Brief history:
The 2-week post-merge CI green-rate watch is on the maintainer
who lands this: ≥ 95 % required per the plan's stability gate;
otherwise pull both jobs and reopen the routing investigation.
- 2026-05-07: 10/10 sweep × 22 tests = 220/220 hard-pass
established a working stability bar after the `:quic`
post-handshake bidi-drop fix landed via `origin/main` (commits
`2a4c07ae`, `d5c854be`, `b622d0c9`, `86a4727e`, `31d19258`).
- Commit `21947bc5` re-added the `hang-interop` +
`browser-interop` jobs to `.github/workflows/build.yml`.
- Maintainer review then deferred CI gating on cost grounds
(cold cache ~10 min hang, ~13 min browser; most PRs don't
touch audio / MoQ / QUIC). Both jobs were removed; the YAML
shape stays preserved in the closed-but-deferred plan
[`2026-05-07-cross-stack-interop-ci-gating.md`](2026-05-07-cross-stack-interop-ci-gating.md)
for re-evaluation if the cost calculus changes.
The trace-capture instrumentation
(`-DnestsHangInteropTraceRelay=true`) added during the routing
investigation stays in place — useful when triaging a future
flake.
## Pending follow-ups
@@ -1,19 +1,26 @@
# Plan: wire CI gating for the cross-stack interop suite
**Status:** ✅ CLOSED 2026-05-07. Both jobs wired in commit
`21947bc5` (path-tweaked from the original removed shape per
the `nestsClient/tests/browser-interop/` move). Stability-bar
sweep: **10/10 BUILD SUCCESSFUL × 22 tests = 220/220 pass.**
Acceptance bar from the roadmap met.
**Status:** ⏸ DEFERRED 2026-05-07.
The infrastructure to wire CI is ready — both jobs landed in
commit `21947bc5` and a 10/10 stability sweep × 22 tests =
220/220 passed on the agent rig. A maintainer review then
judged the wallclock cost too high for the change-pattern and
removed the jobs (cold cache: ~10 min hang, ~13 min browser;
most PRs don't touch audio / MoQ / QUIC).
The suites are now run manually before merging changes that
touch the relevant code paths. Developer-facing docs:
[`../tests/README.md`](../tests/README.md). This plan stays
in the tree for the next time someone wants to revisit CI
gating — the YAML shapes below are the exact reverse-merge
target, and the stability bar (10/10 sweep) has already been
demonstrated on this branch.
**Depended on:**
- `2026-05-07-moq-relay-routing-investigation.md` closed (✅)
- `2026-05-07-tighten-cross-stack-assertions.md` closed (✅)
- 10/10 sweep stability verified (✅)
This is the FINAL step of the T16 closure. With stable hard-pass
suites, CI gating becomes safe and meaningful.
## What's needed
### A) `.github/workflows/build.yml` — the hang-interop job
@@ -67,22 +67,23 @@ HangInteropTest with their current soft-pass assertions intact.
**Acceptance bar met.** 5/5 sweep with hard assertions.
## Priority 3 — `2026-05-07-cross-stack-interop-ci-gating.md` ✅ CLOSED
## Priority 3 — `2026-05-07-cross-stack-interop-ci-gating.md` ⏸ DEFERRED (manual run only)
> **Closed 2026-05-07** by commit `21947bc5`. Both
> `hang-interop` and `browser-interop` jobs landed in
> `.github/workflows/build.yml`, gated on `lint`,
> `ubuntu-latest`, 30 min timeout, with cached cargo + bun +
> Playwright Chromium. Path-tweaked from the original removed
> shape because the browser harness moved from
> `nestsClient-browser-interop/` to
> `nestsClient/tests/browser-interop/` (commit `bd7b166f`).
**Acceptance bar met.** 10/10 sweep BUILD SUCCESSFUL × 22 tests
= **220/220 pass** (~5m 28s steady state per sweep on the agent
rig). The first 2 weeks of post-merge CI runs still need a
maintainer to monitor flake rate (per the plan's "≥ 95% green
rate" gate); pull the jobs again if it falls below.
> **Briefly wired then reverted, 2026-05-07.** Commit `21947bc5`
> re-added both jobs to `build.yml`; verification gave 10/10
> sweep × 22 tests = 220/220 hard-pass on the agent rig. A
> subsequent maintainer review judged the cold-cache cost
> (~10 min hang, ~13 min browser) too high for the change-pattern
> (most PRs don't touch audio / MoQ / QUIC) and removed the jobs.
>
> The suites stay green locally and are intended to run before
> merging any change that touches the audio / moq-lite / `:quic`
> stack. Documentation: [`nestsClient/tests/README.md`](../tests/README.md)
> covers when/how/prerequisites/debugging.
>
> The CI-gating plan is kept around in case the wallclock cost
> calculus changes (e.g., a much faster runner, or a smaller
> opt-in subset of the suite that runs in <2 min).
## Independent track — `2026-05-07-framespergroup-production-rerun.md`