open-standards/.github/workflows/wai-sink-android.yml
Transaction Science 6517401f01 CI: the gate that decides whether a workflow runs was queued behind the work it gates
Every path-gated workflow begins with one small job -- `does this change touch
it`, or `which workspaces this change touches` for the standards build -- that
all of its real jobs declare in `needs`. All 55 of those gate jobs ran on
`linux-arm64`.

That is the one label the fleet cannot spare. Sampled while 267 runs were
queued, every waiting run had the same shape:

    [waiting]  linux-arm64    does this change touch it
    [blocked]  linux          member · dataflow device, simulated
    [blocked]  macos          member · dataflow device absent

237 runs, each holding a gate in the longest queue, and every real job blocked
behind it. Not only `linux` jobs: `macos`, `linux-hw`, `linux-arm64-hw` and
`macos-m3-ultra` jobs were blocked too. The measured effect was `linux-arm64`
263 deep against `linux` **zero**, with jetson-hub's 20 cores taking no job for
a full hour -- starved not for want of work but because the work could not be
unblocked. A gate waiting on the resource it is rationing is head-of-line
blocking, and it makes every other runner in the fleet idle on its behalf.

The gates move to `linux`. Nothing else changes: the diff is 55 `runs-on` lines
and no others, and every touched file still parses.

Why `linux` and not elsewhere. The gate is not free -- it compares
`BASE...HEAD`, which needs the merge base, so `fetch-depth: 0` is a full clone
rather than a ten-second script. `macos` is served by hosts on a Mac Studio
sitting at load 185; `linux` is a separate 20-core box that was measurably
empty. This moves a serialisation point to the least contended label; it does
not remove it, and the gate is still a single point of failure for its
workflow. The structural fix is an early-exit first step inside each job, which
costs one checkout per leg and is a larger change.

Follow-up, deliberately not bundled: jetson-hub runs at capacity 2 with 15 GB of
RAM, so 55 clones still serialise two at a time. Raising capacity helps gates
and hurts the heavy emulated-architecture jobs that share the label, so it wants
measuring on its own.

Co-Authored-By: Claude Opus 5 <[email protected]>
2026-11-03 15:19:13 -05:00

167 lines
7.3 KiB
YAML

name: wai-sink-android native sink library
# The native sink's Kotlin library (wai/wai-sink-android, extensions/native-sink.md).
#
# On the linux-arm64 runner, the platform-free half's unit tests on the host JVM: libwai and the
# JNI layer built for the runner, then the corpus live-conformance/native-sink through the JNI
# layer, every case and every refused configuration; the page-parity bundle byte for byte; the
# C interface's statuses as the JNI layer carries them; the loopback relay over the reference
# package's floor (presented byte for byte, a changed fragment answered 502 on every request,
# ranges, content codings, an unsigned playlist refused, nothing answered outside the session's
# token); and the observation loop's hidden and idle endings with its clock advanced by the test.
# No Android SDK is used there.
#
# On the macOS runner, the host where the platform SDK, its NDK and its emulator are installed:
# libwai and the JNI layer for each ABI with the pinned NDK, the AAR, and the instrumented tests in
# a headless emulator (the host's ABI, arm64-v8a): the corpus and the relay on the device, the
# platform's media player library playing the floor through the sink to its end, a changed
# fragment refused on every retry, a session nobody can see ended as hidden, and the attention
# signals read from a real activity. Nothing listens beyond loopback.
#
# Push filters match the push's delta, so every path the jobs read is listed.
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
on:
push:
branches: [main]
paths:
- 'wai/wai-sink-android/**'
- 'wai/wai-rs/**'
- 'wai/wai-quantum/**'
- 'wai/wai-ops/**'
- 'wai/live-conformance/native-sink/**'
- 'wai/extensions/native-sink.md'
- 'rust-toolchain.toml'
- '.github/workflows/wai-sink-android.yml'
pull_request:
paths:
- 'wai/wai-sink-android/**'
- 'wai/wai-rs/**'
- 'wai/wai-quantum/**'
- 'wai/wai-ops/**'
- 'wai/live-conformance/native-sink/**'
- 'wai/extensions/native-sink.md'
- 'rust-toolchain.toml'
- '.github/workflows/wai-sink-android.yml'
jobs:
relevant:
name: does this change touch it
runs-on: linux
outputs:
run: ${{ steps.gate.outputs.run }}
steps:
- uses: actions/checkout@v5
with:
fetch-depth: 0
- id: gate
run: |
set -euo pipefail
r=$(python3 scripts/pr-touches-paths.py .github/workflows/wai-sink-android.yml \
"${{ github.event.pull_request.base.sha }}" \
"${{ github.event.pull_request.head.sha }}")
echo "this change touches wai-sink-android: $r"
echo "run=$r" >> "$GITHUB_OUTPUT"
jvm-linux-arm64:
name: the JNI layer and the platform-free half on the host JVM · linux-arm64
needs: relevant
if: needs.relevant.outputs.run == 'true'
runs-on: linux-arm64
timeout-minutes: 60
steps:
- uses: actions/checkout@v5
- name: Install Rust 1.98.0
shell: bash
run: |
export PATH="$HOME/.cargo/bin:$PATH"
command -v rustup >/dev/null 2>&1 || \
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --profile minimal --default-toolchain 1.98.0
rustup toolchain install 1.98.0 --profile minimal
rustup default 1.98.0
echo "$HOME/.cargo/bin" >> "$GITHUB_PATH"
# A JDK 17, from the image's packages when the image has none.
- name: JDK 17
shell: bash
run: |
set -euo pipefail
j="/usr/lib/jvm/java-17-openjdk-$(dpkg --print-architecture)"
if [ ! -x "$j/bin/javac" ]; then
sudo=$(command -v sudo || true)
$sudo apt-get update -qq
$sudo apt-get install -y -qq --no-install-recommends openjdk-17-jdk-headless >/dev/null
fi
test -f "$j/include/jni.h" || { echo "::error::no JDK 17 with jni.h at $j"; exit 1; }
echo "JAVA_HOME=$j" >> "$GITHUB_ENV"
- name: libwai and the JNI layer for this runner
shell: bash
working-directory: wai/wai-sink-android
run: bash scripts/build-native.sh host
# The pinned Gradle and the build's dependencies are kept on the runner's cache volume when
# it has one.
- name: The unit tests
shell: bash
working-directory: wai/wai-sink-android
run: |
set -euo pipefail
cache=/ci-cache/wai-sink-android
mkdir -p "$cache" 2>/dev/null || cache="$RUNNER_TEMP/wai-sink-android"
export WAI_SINK_GRADLE_DISTS="$cache/dists" GRADLE_USER_HOME="$cache/gradle-home"
bash scripts/gradle.sh -Pwai.sink.jvmOnly=true :core:test
python3 -I - core/build/test-results/test <<'PY'
import sys, pathlib, xml.etree.ElementTree as ET
runs = {}
tests = fails = skips = 0
for f in pathlib.Path(sys.argv[1]).glob("*.xml"):
s = ET.parse(f).getroot()
tests += int(s.get("tests", 0)); fails += int(s.get("failures", 0)) + int(s.get("errors", 0)); skips += int(s.get("skipped", 0))
runs[s.get("name").split(".")[-1]] = int(s.get("tests", 0))
print(f"unit tests: {tests} run, {fails} failed, {skips} skipped: {runs}")
need = {"CorpusTest", "RelayTest", "SessionDriverTest", "JsonTest"}
missing = sorted(c for c in need if runs.get(c, 0) == 0)
if missing: print(f"::error::these classes ran no test: {missing}")
sys.exit(0 if tests > 0 and fails == 0 and skips == 0 and not missing else 1)
PY
android-macos:
name: every ABI, the AAR, and the instrumented tests in a headless emulator · macos
needs: relevant
if: needs.relevant.outputs.run == 'true'
runs-on: macos
timeout-minutes: 90
steps:
- uses: actions/checkout@v5
# launchd's PATH omits cargo on the macOS host runner.
- name: Put cargo on PATH
shell: bash
run: echo "$HOME/.cargo/bin" >> "$GITHUB_PATH"
- name: The targets the ABIs need
shell: bash
run: rustup target add aarch64-linux-android armv7-linux-androideabi x86_64-linux-android
- name: libwai and the JNI layer for each ABI
shell: bash
working-directory: wai/wai-sink-android
run: nice -n 10 bash scripts/build-native.sh android
- name: The AAR
shell: bash
working-directory: wai/wai-sink-android
run: |
set -euo pipefail
nice -n 10 bash scripts/gradle.sh :sink:assembleRelease
aar=sink/build/outputs/aar/sink-release.aar
# The listing is read whole, then searched: under pipefail, `unzip -l | grep -q` fails
# whenever grep stops at its first match before unzip has written the rest (SIGPIPE).
list="$(unzip -l "$aar")"
for abi in arm64-v8a armeabi-v7a x86_64; do
for so in libwai.so libwai_sink_jni.so; do
grep -q " jni/$abi/$so\$" <<<"$list" || { echo "::error::the AAR has no jni/$abi/$so"; echo "$list"; exit 1; }
done
done
grep -q ' classes.jar$' <<<"$list" || { echo "::error::the AAR has no classes"; echo "$list"; exit 1; }
- name: The instrumented tests in a headless emulator
shell: bash
working-directory: wai/wai-sink-android
run: nice -n 10 bash scripts/test-emulator.sh