Files
pplus-core/.gitea/workflows/test.yml
T
alexm b10a5c87a3
Test / gate (push) Canceled after 0s
A gate that does not depend on its machine, on a runner of its own
The gate first ran on the server that deploys the site: a slow container
with no C compiler and no ICU. Ten tests failed there on correct code,
and no Mojo program could be built at all.

- The workflows ask for a runner labeled heavy, not the deploy server:
  the gate builds and runs some 300 Mojo programs and fuzzes. docs/ci.md
  says how to register one. With none registered a run waits.
- The pinned environment has a C compiler on Linux (gcc, with its C
  library pinned to glibc 2.34): mojo build links with the C compiler it
  finds, and conda-forge's default links against a C library too old
  for Mojo's runtime. mojo-local reports a program that could not be
  linked as that, and does not keep the answer.
- The pinned Dafny runs without ICU, and dafny-fetch fails when the
  Dafny it fetched does not start.
- A test's time limit is processor time at the machine's pace
  (tests/helpers/pace.mjs).
- The prover's time on the clock is the caller's: provePseudo and
  Workspace.proofBudgetMs take it, the Prove tab's 3 seconds stay the
  default, and the command line, library-check and the tests use none,
  so their answers are the counted budgets' alone.
- The library pointer moves to its workflow change (e5ae090).
2026-11-13 06:13:28 -05:00

89 lines
4.4 KiB
YAML

# pplus-core's gate on every push and pull request: `pixi run ci`, the same task a developer runs, in the
# environment pixi.lock pins (Node, Python, the Mojo compiler), so CI and a developer cannot drift. docs/ci.md says
# what it runs, how long it takes, and what to do when it fails.
#
# Gitea Actions, on the runner with the label `heavy`: a machine with cores to spare, not the server that deploys
# the site, because the gate builds and runs some 300 Mojo programs and fuzzes (docs/ci.md has the runner's setup).
# With no such runner registered a run waits; nothing here runs on the deploy server. What the job needs beyond git,
# Node and curl comes from pixi.lock or is fetched by a pinned hash, so the runner may be a container or a host.
# The same file runs on GitHub Actions with one change: `runs-on: ubuntu-latest`.
name: Test
on:
push:
pull_request:
workflow_dispatch:
concurrency:
group: test-${{ github.ref }}
cancel-in-progress: true
env:
# set by Gitea and GitHub Actions already; written here so the tests' "fail, not skip, under CI" never depends on it
CI: "true"
# the pixi that reads pixi.lock; when the runner has another version, this one is fetched into the cache below
PIXI_VERSION: "0.83.0"
jobs:
gate:
# a pull request runs only from a branch of this repository: a fork's workflow would run its own code on the
# runner (docs/ci.md). Drop this line for a runner that is a throwaway machine.
if: github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository
runs-on: heavy
timeout-minutes: 60
steps:
# library/ is pplus-library, a submodule pinned by commit: the programs, the corpus, the routines
- uses: actions/checkout@v4
with:
submodules: recursive
# pixi: the image's when it is the pinned version, else the pinned release, fetched into a cache directory
# (which lasts only where the runner keeps a home between jobs) and checked by SHA-256
- name: pixi, pinned
shell: bash
run: |
set -euo pipefail
if command -v pixi >/dev/null && [ "$(pixi --version | cut -d' ' -f2)" = "$PIXI_VERSION" ]; then
echo "pixi $PIXI_VERSION is on the PATH: $(command -v pixi)"; exit 0
fi
dir="${PPLUS_CI_CACHE:-$HOME/.cache/pplus-ci}/pixi-$PIXI_VERSION"
if [ ! -x "$dir/pixi" ]; then
case "$(uname -s)-$(uname -m)" in
Linux-x86_64) asset=pixi-x86_64-unknown-linux-musl.tar.gz; sum=064f371e692aaa3400ac581c89da2843c91437f20c2d9155395a5d0037f33551 ;;
Linux-aarch64) asset=pixi-aarch64-unknown-linux-musl.tar.gz; sum=d08403d8f048626ec93ba7b7c56e3b747a5f3a95c4e6e0c815ceb264f5f0b9ce ;;
*) echo "no pinned pixi for $(uname -sm)" >&2; exit 1 ;;
esac
tmp="$(mktemp -d)"
curl -fsSL --retry 3 -o "$tmp/$asset" "https://github.com/prefix-dev/pixi/releases/download/v$PIXI_VERSION/$asset"
echo "$sum $tmp/$asset" | sha256sum -c -
mkdir -p "$dir" && tar -xzf "$tmp/$asset" -C "$dir" pixi && rm -rf "$tmp"
fi
echo "$dir" >> "$GITHUB_PATH"
"$dir/pixi" --version
- name: The pinned environment, exactly as pixi.lock has it
run: pixi install --locked
# .tools/ holds the pinned Dafny and mojo-local's answers (keyed by compiler, arguments and source); it is a
# link into the cache directory, so where the runner keeps a home between jobs neither is fetched or built again
- name: The pinned Dafny
shell: bash
run: |
set -euo pipefail
cache="${PPLUS_CI_CACHE:-$HOME/.cache/pplus-ci}/core-tools"
mkdir -p "$cache"
if [ -d .tools ] && [ ! -L .tools ]; then echo ".tools is a directory in this checkout already: using it"; else ln -sfn "$cache" .tools; fi
pixi run dafny-fetch
# every test, mojo-status-strict, library-check, mojo-local and a short mojo-fuzz and prove-fuzz (pixi.toml)
- name: pixi run ci
run: pixi run ci
# the gate rewrites what it records (the library's status.json and STATUS.md, the corpus's STATUS.md); a
# difference means a change was committed without what it regenerates
- name: Nothing the gate regenerates differs from what is committed
shell: bash
run: |
set -euo pipefail
git diff --stat --exit-code
git -C library diff --stat --exit-code