open-standards/.github/workflows/wai-quantum-formats.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

174 lines
8.6 KiB
YAML

name: wai-quantum other formats
# The readers and writers of other formats (wai/wai-quantum/src/engine/formats/):
# QIR under its base and adaptive profiles and jeff through the crate's own
# Cap'n Proto reader and writer. Their unit tests (the wire format, the
# OpenQASM they write, every intrinsic and gate held to its matrix, each
# profile's refusals with their lines, the Pauli-product rotation against
# exp(iθP), every prefix of a module refused), and the conformance set
# (wai/quantum-conformance/formats/: the QIR specification's examples, the
# jeff project's benchmarks and examples, hand-built modules and OpenQASM
# sources written to each format, each with what the engine must do with it,
# the QIR written pinned by its BLAKE3), and hostile inputs: every file
# of the set mutated, each converted or refused with a code within its
# bounds, natively and under wasm32-wasip2.
#
# The set runs natively, where a second test holds its bundle equal to its
# files, and under wasm32-wasip2 by wasmtime, where the cases come from the
# bundle the test compiles in and the QIR written must be the same bytes.
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
on:
push:
branches: [main]
paths:
- 'wai/wai-quantum/**'
- 'wai/quantum-conformance/formats/**'
- 'rust-toolchain.toml'
- '.github/workflows/wai-quantum-formats.yml'
pull_request:
paths:
- 'wai/wai-quantum/**'
- 'wai/quantum-conformance/formats/**'
- 'rust-toolchain.toml'
- '.github/workflows/wai-quantum-formats.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-quantum-formats.yml \
"${{ github.event.pull_request.base.sha }}" \
"${{ github.event.pull_request.head.sha }}")
echo "this change touches the readers of other formats: $r"
echo "run=$r" >> "$GITHUB_OUTPUT"
native:
name: readers, writers and the set · linux
needs: relevant
if: needs.relevant.outputs.run == 'true'
runs-on: linux
timeout-minutes: 60
steps:
- uses: actions/checkout@v5
- name: Install Rust 1.98.0
shell: bash
run: |
export PATH="$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" >> "$GITHUB_PATH"
# A module filter is satisfied by any one test of the module, so each
# module's tests are required by name to have run.
- name: The readers' and writers' unit tests
shell: bash
working-directory: wai/wai-quantum
run: |
set -euo pipefail
cargo test --locked --release --features full --lib engine::formats 2>&1 | tee "$RUNNER_TEMP/formats.log"
grep -qE '^test result: ok\. [1-9][0-9]+ passed; 0 failed' "$RUNNER_TEMP/formats.log" || { echo "::error::the formats' tests did not pass"; exit 1; }
for m in engine::formats::capnp::tests engine::formats::emit::tests engine::formats::tests::every_gate_is_written_exactly engine::formats::tests::a_pauli_product_rotation_is_exp_i_theta_p engine::formats::tests::the_conformance_set_holds_these_modules engine::formats::tests::a_register_too_large_is_refused_before_it_is_allocated; do
grep -q "^test $m.* \.\.\. ok$" "$RUNNER_TEMP/formats.log" || { echo "::error::no test of $m ran"; exit 1; }
done
- name: The conformance set
shell: bash
working-directory: wai/wai-quantum
run: |
set -euo pipefail
cargo test --locked --release --features full --test formats_conformance 2>&1 | tee "$RUNNER_TEMP/set.log"
grep -q 'test result: ok. 2 passed; 0 failed' "$RUNNER_TEMP/set.log" || { echo "::error::the conformance set did not pass"; exit 1; }
# Hostile inputs: every file of the set mutated by a seeded generator
# (bits flipped, files cut short and spliced, and every length, count,
# offset and 32-bit field inflated), each converted or refused with a
# code within its time and memory bounds. The count of inputs is pinned.
- name: Hostile inputs
shell: bash
working-directory: wai/wai-quantum
run: |
set -euo pipefail
cargo test --locked --release --features full --test formats_mutation -- --nocapture 2>&1 | tee "$RUNNER_TEMP/mutation.log"
grep -q 'test result: ok. 1 passed; 0 failed' "$RUNNER_TEMP/mutation.log" || { echo "::error::a hostile input was not refused cleanly"; exit 1; }
grep -q '45153 inputs: ' "$RUNNER_TEMP/mutation.log" || { echo "::error::the count of hostile inputs is not the pinned 45153"; exit 1; }
- name: The readers without the receipts
shell: bash
working-directory: wai/wai-quantum
run: |
set -euo pipefail
cargo check --locked --no-default-features --features quantum_engine --lib
wasip2:
name: the set under wasm32-wasip2 (wasmtime)
needs: relevant
if: needs.relevant.outputs.run == 'true'
runs-on: linux
timeout-minutes: 60
steps:
- uses: actions/checkout@v5
- name: Install Rust 1.98.0 with the wasm32-wasip2 target
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
rustup target add --toolchain 1.98.0 wasm32-wasip2
echo "$HOME/.cargo/bin" >> "$GITHUB_PATH"
# The release binary, pinned by version and SHA-256, rather than a build from source.
- name: Install wasmtime 49.0.2
shell: bash
run: |
set -euo pipefail
[ "$(uname -m)" = x86_64 ] || { echo "::error::the pinned wasmtime archive is x86_64; this host is $(uname -m)"; exit 1; }
if ! command -v xz >/dev/null 2>&1; then
SUDO=""; [ "$(id -u)" -ne 0 ] && SUDO="sudo"
$SUDO apt-get update -qq
$SUDO apt-get install -y -qq xz-utils
fi
v=49.0.2
f="wasmtime-v$v-x86_64-linux.tar.xz"
curl --proto '=https' --tlsv1.2 -sSfL -o "$RUNNER_TEMP/$f" \
"https://github.com/bytecodealliance/wasmtime/releases/download/v$v/$f"
echo "a4d6e9e3a5a60f527cf7793d674c48930c80c2e8977995b8a275cad3254b9322 $RUNNER_TEMP/$f" | sha256sum -c -
tar -xJf "$RUNNER_TEMP/$f" -C "$RUNNER_TEMP"
echo "$RUNNER_TEMP/wasmtime-v$v-x86_64-linux" >> "$GITHUB_PATH"
# A debug build, so integer overflow panics instead of wrapping. The test
# that reads the set's files has no wasip2 build; the one that runs the
# cases from the compiled-in bundle does.
- name: The conformance set under wasmtime
shell: bash
working-directory: wai/wai-quantum
env:
CARGO_TARGET_WASM32_WASIP2_RUNNER: wasmtime
run: |
set -euo pipefail
cargo test --locked --target wasm32-wasip2 --features full --test formats_conformance 2>&1 | tee "$RUNNER_TEMP/wasip2.log"
grep -q 'test result: ok. 1 passed; 0 failed' "$RUNNER_TEMP/wasip2.log" || { echo "::error::the conformance set did not pass under wasip2"; exit 1; }
# The same hostile inputs where an allocation larger than the address
# space panics: a release build with overflow checks on, as it is slow
# enough otherwise to outlast the job.
- name: Hostile inputs under wasmtime
shell: bash
working-directory: wai/wai-quantum
env:
CARGO_TARGET_WASM32_WASIP2_RUNNER: wasmtime
CARGO_PROFILE_RELEASE_OVERFLOW_CHECKS: "true"
run: |
set -euo pipefail
cargo test --locked --release --target wasm32-wasip2 --features full --test formats_mutation -- --nocapture 2>&1 | tee "$RUNNER_TEMP/wasip2-mutation.log"
grep -q 'test result: ok. 1 passed; 0 failed' "$RUNNER_TEMP/wasip2-mutation.log" || { echo "::error::a hostile input was not refused cleanly under wasip2"; exit 1; }
grep -q '45153 inputs: ' "$RUNNER_TEMP/wasip2-mutation.log" || { echo "::error::the count of hostile inputs under wasip2 is not the pinned 45153"; exit 1; }