ps5upload/.github/workflows/docker-engine.yml

name: docker-engine

# Builds and publishes two GHCR images from the engine crate:
#   - ps5upload-engine        (engine/Dockerfile)       — minimal, no UI
#   - ps5upload-engine-webui  (engine/Dockerfile.webui) — engine + the full
#     React app served over HTTP, for browser/NAS/headless use (no Tauri
#     client needed). Its build context is the REPO ROOT, not engine/, since
#     it needs to `npm ci && npm run build:vite` in client/ before the Rust
#     build embeds the output.
#
# Tag pushes (v*) publish multi-arch (amd64 + arm64) manifests for both
# images, in lockstep — one release cuts both. PRs that touch engine/ or
# client/ get a build-only smoke-check of both Dockerfiles (no push).
#
# Both images embed the two PS5 ELF images (ps5upload.elf and the DPI
# install daemon) so a browser driving a self-hosted engine can bring the
# DPI daemon up and put the helper back afterwards — the fallback that
# installs a game PATCH. The desktop client does this from its own embeds;
# a browser has no socket and no bytes, so without them the web UI could
# install base games and nothing else (the web UI half of #152). That is why the tag
# builds now start with a `build-payload` job: the main payload is not
# committed (see .gitignore) and needs the PS5 payload SDK.
#
# Manual re-run (e.g. after an Actions outage wedges a tag-push run):
#   gh workflow run docker-engine.yml -f tag=v3.3.16
on:
  push:
    tags: ['v*']
  pull_request:
    paths: ['engine/**', 'client/**', 'payload/**', 'scripts/docker-smoke.sh', '.github/workflows/docker-engine.yml']
  workflow_dispatch:
    inputs:
      tag:
        description: "Release tag to publish (e.g. v3.3.16). Must already exist."
        required: true

# Least privilege at the top: read-only by default. Only the jobs that actually
# push to GHCR (build*, merge*) opt into packages:write below — the PR
# `verify`/`verify-webui` jobs inherit read-only and never get a write-capable
# token.
permissions:
  contents: read

concurrency:
  group: docker-engine-${{ github.event.inputs.tag || github.ref }}
  cancel-in-progress: true

jobs:
  # ── PR gate: build amd64, no push ──────────────────────────────────────────
  verify:
    if: github.event_name == 'pull_request'
    runs-on: ubuntu-24.04
    # A full engine (and, for the web UI, client) build: ~7-10 min on a good
    # runner, so 10 cancelled ordinary runs (PR #338).
    timeout-minutes: 25
    steps:
      - uses: actions/checkout@v7

      - name: Stage the committed DPI daemon for the embed
        # ps5upload.elf is not in the repo (it needs the payload SDK), so a
        # PR build can only cover the DPI half. That is the half the web
        # UI's patch install depends on, and staging it here means a broken
        # embed fails the PR rather than the release.
        run: |
          set -euo pipefail
          mkdir -p engine/crates/ps5upload-engine/payload
          cp -v payload/installer/ps5upload-installer.elf engine/crates/ps5upload-engine/payload/

      - uses: docker/setup-buildx-action@v4

      - uses: docker/build-push-action@v7
        with:
          context: .
          file: engine/Dockerfile
          platforms: linux/amd64
          push: false
          # Load it so the smoke test below can run it.
          load: true
          tags: ps5upload-engine:smoke
          cache-from: type=gha
          cache-to: type=gha,mode=min

      # Building proved nothing about whether the image works: 5.40.0 built
      # fine and refused every request, had no /tmp for uploads and no
      # writable home. Run it and check those.
      - name: Smoke-test the image
        run: scripts/docker-smoke.sh ps5upload-engine:smoke

  # ── PR gate for the webui image: same idea, root context + Dockerfile.webui ─
  verify-webui:
    if: github.event_name == 'pull_request'
    runs-on: ubuntu-24.04
    # A full engine (and, for the web UI, client) build: ~7-10 min on a good
    # runner, so 10 cancelled ordinary runs (PR #338).
    timeout-minutes: 25
    steps:
      - uses: actions/checkout@v7

      - name: Stage the committed DPI daemon for the embed
        # ps5upload.elf is not in the repo (it needs the payload SDK), so a
        # PR build can only cover the DPI half. That is the half the web
        # UI's patch install depends on, and staging it here means a broken
        # embed fails the PR rather than the release.
        run: |
          set -euo pipefail
          mkdir -p engine/crates/ps5upload-engine/payload
          cp -v payload/installer/ps5upload-installer.elf engine/crates/ps5upload-engine/payload/

      - uses: docker/setup-buildx-action@v4

      - uses: docker/build-push-action@v7
        with:
          context: .
          file: engine/Dockerfile.webui
          platforms: linux/amd64
          push: false
          load: true
          tags: ps5upload-engine-webui:smoke
          cache-from: type=gha,scope=webui
          cache-to: type=gha,mode=min,scope=webui

      - name: Smoke-test the image
        run: scripts/docker-smoke.sh ps5upload-engine-webui:smoke --webui

  # ── Resolve + validate the tag ONCE, hand the trusted values to build/merge ─
  # Centralizing this means the tag is regex-validated before it ever reaches a
  # checkout `ref:` (no ref injection from a hand-typed workflow_dispatch input),
  # and the lowercased image names are computed once via the safe env pattern
  # instead of interpolating ${{ github.repository_owner }} into shell.
  resolve:
    if: github.event_name != 'pull_request'
    runs-on: ubuntu-24.04
    timeout-minutes: 5
    outputs:
      tag: ${{ steps.t.outputs.tag }}
      version: ${{ steps.t.outputs.version }}
      major_minor: ${{ steps.t.outputs.major_minor }}
      image: ${{ steps.t.outputs.image }}
      image_webui: ${{ steps.t.outputs.image_webui }}
    steps:
      - name: Validate tag + derive image names
        id: t
        env:
          INPUT_TAG: ${{ github.event.inputs.tag || github.ref_name }}
          OWNER: ${{ github.repository_owner }}
        run: |
          raw_tag="$INPUT_TAG"
          if ! printf '%s' "$raw_tag" | grep -Eq '^v[0-9]+\.[0-9]+\.[0-9]+([-+._A-Za-z0-9]*)?$'; then
            echo "::error::Refusing to proceed: '$raw_tag' is not a valid version tag"
            exit 1
          fi
          version="${raw_tag#v}"
          owner_lc="$(printf '%s' "$OWNER" | tr '[:upper:]' '[:lower:]')"
          {
            echo "tag=$raw_tag"
            echo "version=$version"
            echo "major_minor=${version%.*}"
            echo "image=ghcr.io/${owner_lc}/ps5upload-engine"
            echo "image_webui=ghcr.io/${owner_lc}/ps5upload-engine-webui"
          } >> "$GITHUB_OUTPUT"

  # ── Build the two PS5 ELF images the engine embeds ─────────────────────────
  # Mirrors publish.yml's job of the same name. Duplicated rather than shared
  # because the two workflows run independently on a tag push and neither can
  # wait on the other's artifacts. The DPI daemon is committed, but the main
  # payload is not (see .gitignore), and the engine needs BOTH: it streams the
  # daemon to the loader on :9021 and then puts ps5upload back afterwards.
  build-payload:
    needs: resolve
    if: github.event_name != 'pull_request'
    runs-on: ubuntu-24.04
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@v7
        with:
          ref: ${{ needs.resolve.outputs.tag }}

      - uses: actions/setup-node@v7
        with:
          node-version: 22
          cache: npm
          cache-dependency-path: client/package-lock.json

      # make payload embeds the console web page (payload/Makefile: build/webui_page.c runs
      # `npm run build:console` in client/), so the client's dependencies must be installed.
      - name: Install client deps (console web page)
        working-directory: client
        run: npm ci --no-audit --no-fund

      - name: Install payload toolchain
        timeout-minutes: 12
        run: |
          set -euo pipefail
          # apt has no default network timeout, so a stalled mirror hangs
          # until the job dies. Retry with explicit timeouts.
          apt_retry() {
            for attempt in 1 2 3; do
              if sudo DEBIAN_FRONTEND=noninteractive apt-get \
                   -o Acquire::Retries=3 \
                   -o Acquire::http::Timeout=20 \
                   -o Acquire::https::Timeout=20 \
                   "$@"; then
                return 0
              fi
              echo "apt attempt $attempt failed; retrying in 15s" >&2
              sleep 15
            done
            echo "apt failed after 3 attempts: apt-get $*" >&2
            return 1
          }
          apt_retry update
          # llvm-22 is not in Ubuntu 24.04's default repos.
          curl -fsSL --retry 3 --retry-delay 5 https://apt.llvm.org/llvm.sh | sudo bash -s -- 22
          apt_retry install -y clang-22 llvm-22-dev lld-22 unzip python3

      - name: Install pinned PS5 Payload SDK
        run: sudo env PS5_SDK_INSTALL_DIR=/opt/ps5-payload-sdk bash scripts/install-ps5-sdk.sh

      - name: Build payload ELFs
        env:
          PS5_PAYLOAD_SDK: /opt/ps5-payload-sdk
          # SDK v0.42's prospero-llvm-config candidate search maxes at 21,
          # so point it at the CI-installed LLVM 22 explicitly.
          LLVM_CONFIG: /usr/bin/llvm-config-22
        run: make payload

      - name: Sanity-check ELF shape
        run: |
          file payload/ps5upload.elf payload/installer/ps5upload-installer.elf
          test -s payload/ps5upload.elf
          test -s payload/installer/ps5upload-installer.elf

      - uses: actions/upload-artifact@v7
        with:
          name: ps5upload-payload-elves
          path: |
            payload/ps5upload.elf
            payload/installer/ps5upload-installer.elf
          if-no-files-found: error
          retention-days: 1

  # ── Per-arch native build, push by digest (no tag yet) ─────────────────────
  build:
    needs: [resolve, build-payload]
    if: github.event_name != 'pull_request'
    runs-on: ${{ matrix.os }}
    timeout-minutes: 30
    permissions:
      contents: read
      packages: write
    strategy:
      fail-fast: false
      matrix:
        include:
          - { os: ubuntu-24.04,     platform: linux/amd64 }
          - { os: ubuntu-24.04-arm, platform: linux/arm64 }
    env:
      IMAGE: ${{ needs.resolve.outputs.image }}
    steps:
      - uses: actions/checkout@v7
        with:
          ref: ${{ needs.resolve.outputs.tag }}

      - name: Prepare environment
        env:
          PLATFORM: ${{ matrix.platform }}
        run: echo "PLATFORM_PAIR=$(printf '%s' "$PLATFORM" | tr '/' '-')" >> "$GITHUB_ENV"

      - name: Stage the PS5 ELF images for the engine embed
        uses: actions/download-artifact@v8
        with:
          name: ps5upload-payload-elves
          path: /tmp/payload-elves

      - name: Move them into the engine's embed directory
        # `build.rs` looks here first — the same convention `webui/` uses for
        # the React bundle, so the images ride along in the existing
        # `COPY engine/` layer and no Dockerfile needs a second context path.
        run: |
          set -euo pipefail
          mkdir -p engine/crates/ps5upload-engine/payload
          cp -v /tmp/payload-elves/ps5upload.elf \
                /tmp/payload-elves/installer/ps5upload-installer.elf \
                engine/crates/ps5upload-engine/payload/
          ls -la engine/crates/ps5upload-engine/payload/

      - uses: docker/setup-buildx-action@v4

      - uses: docker/login-action@v4.6.0
        with:
          registry: ghcr.io
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}

      - uses: docker/metadata-action@v6
        id: meta
        with:
          images: ${{ env.IMAGE }}

      # Same build, loaded locally and run before anything is pushed: a
      # release tag can't be moved, so a broken image would need a new release.
      - name: Build for the smoke test
        uses: docker/build-push-action@v7
        with:
          context: .
          file: engine/Dockerfile
          platforms: ${{ matrix.platform }}
          load: true
          tags: ps5upload-smoke:local
          cache-from: type=registry,ref=${{ env.IMAGE }}:buildcache-${{ env.PLATFORM_PAIR }}

      - name: Smoke-test the image
        run: scripts/docker-smoke.sh ps5upload-smoke:local

      - name: Build and push by digest
        id: push
        uses: docker/build-push-action@v7
        with:
          context: .
          file: engine/Dockerfile
          platforms: ${{ matrix.platform }}
          labels: ${{ steps.meta.outputs.labels }}
          outputs: type=image,name=${{ env.IMAGE }},push-by-digest=true,name-canonical=true,push=true
          cache-from: type=registry,ref=${{ env.IMAGE }}:buildcache-${{ env.PLATFORM_PAIR }}
          cache-to: type=registry,ref=${{ env.IMAGE }}:buildcache-${{ env.PLATFORM_PAIR }},mode=max

      - name: Export digest
        env:
          DIGEST: ${{ steps.push.outputs.digest }}
        run: |
          mkdir -p /tmp/digests
          touch "/tmp/digests/${DIGEST#sha256:}"

      - uses: actions/upload-artifact@v7
        with:
          name: digests-${{ env.PLATFORM_PAIR }}
          path: /tmp/digests/*
          if-no-files-found: error
          retention-days: 1

  # ── Per-arch native build of the webui image, push by digest ───────────────
  # Mirrors `build` above; only the context/file/image/artifact-name/cache-
  # scope differ. Kept as a separate job (not a matrix dimension on `build`)
  # so a webui-only failure doesn't need `fail-fast: false` gymnastics to
  # keep the plain engine image publishing on time.
  build-webui:
    needs: [resolve, build-payload]
    if: github.event_name != 'pull_request'
    runs-on: ${{ matrix.os }}
    timeout-minutes: 30
    permissions:
      contents: read
      packages: write
    strategy:
      fail-fast: false
      matrix:
        include:
          - { os: ubuntu-24.04,     platform: linux/amd64 }
          - { os: ubuntu-24.04-arm, platform: linux/arm64 }
    env:
      IMAGE: ${{ needs.resolve.outputs.image_webui }}
    steps:
      - uses: actions/checkout@v7
        with:
          ref: ${{ needs.resolve.outputs.tag }}

      - name: Prepare environment
        env:
          PLATFORM: ${{ matrix.platform }}
        run: echo "PLATFORM_PAIR=$(printf '%s' "$PLATFORM" | tr '/' '-')" >> "$GITHUB_ENV"

      - name: Stage the PS5 ELF images for the engine embed
        uses: actions/download-artifact@v8
        with:
          name: ps5upload-payload-elves
          path: /tmp/payload-elves

      - name: Move them into the engine's embed directory
        # `build.rs` looks here first — the same convention `webui/` uses for
        # the React bundle, so the images ride along in the existing
        # `COPY engine/` layer and no Dockerfile needs a second context path.
        run: |
          set -euo pipefail
          mkdir -p engine/crates/ps5upload-engine/payload
          cp -v /tmp/payload-elves/ps5upload.elf \
                /tmp/payload-elves/installer/ps5upload-installer.elf \
                engine/crates/ps5upload-engine/payload/
          ls -la engine/crates/ps5upload-engine/payload/

      - uses: docker/setup-buildx-action@v4

      - uses: docker/login-action@v4.6.0
        with:
          registry: ghcr.io
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}

      - uses: docker/metadata-action@v6
        id: meta
        with:
          images: ${{ env.IMAGE }}

      # Same build, loaded locally and run before anything is pushed: a
      # release tag can't be moved, so a broken image would need a new release.
      - name: Build for the smoke test
        uses: docker/build-push-action@v7
        with:
          context: .
          file: engine/Dockerfile.webui
          platforms: ${{ matrix.platform }}
          load: true
          tags: ps5upload-smoke:local
          cache-from: type=registry,ref=${{ env.IMAGE }}:buildcache-${{ env.PLATFORM_PAIR }}

      - name: Smoke-test the image
        run: scripts/docker-smoke.sh ps5upload-smoke:local --webui

      - name: Build and push by digest
        id: push
        uses: docker/build-push-action@v7
        with:
          context: .
          file: engine/Dockerfile.webui
          platforms: ${{ matrix.platform }}
          labels: ${{ steps.meta.outputs.labels }}
          outputs: type=image,name=${{ env.IMAGE }},push-by-digest=true,name-canonical=true,push=true
          cache-from: type=registry,ref=${{ env.IMAGE }}:buildcache-${{ env.PLATFORM_PAIR }}
          cache-to: type=registry,ref=${{ env.IMAGE }}:buildcache-${{ env.PLATFORM_PAIR }},mode=max

      - name: Export digest
        env:
          DIGEST: ${{ steps.push.outputs.digest }}
        run: |
          mkdir -p /tmp/digests
          touch "/tmp/digests/${DIGEST#sha256:}"

      # Distinct "webui-digests-" prefix (not just a "webui-" infix) so the
      # plain engine `merge` job's `pattern: digests-*` can't glob-match
      # these artifacts and fold a webui digest into the minimal image's
      # manifest.
      - uses: actions/upload-artifact@v7
        with:
          name: webui-digests-${{ env.PLATFORM_PAIR }}
          path: /tmp/digests/*
          if-no-files-found: error
          retention-days: 1

  # ── Merge per-arch digests into a multi-arch manifest ──────────────────────
  merge:
    needs: [resolve, build]
    if: github.event_name != 'pull_request'
    runs-on: ubuntu-24.04
    timeout-minutes: 10
    permissions:
      contents: read
      packages: write
    env:
      IMAGE: ${{ needs.resolve.outputs.image }}
      VERSION: ${{ needs.resolve.outputs.version }}
      MAJOR_MINOR: ${{ needs.resolve.outputs.major_minor }}
    steps:
      - uses: actions/download-artifact@v8
        with:
          path: /tmp/digests
          pattern: digests-*
          merge-multiple: true

      - uses: docker/setup-buildx-action@v4

      - uses: docker/login-action@v4.6.0
        with:
          registry: ghcr.io
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}

      - uses: docker/metadata-action@v6
        id: meta
        with:
          images: ${{ env.IMAGE }}
          tags: |
            type=raw,value=${{ env.VERSION }}
            type=raw,value=${{ env.MAJOR_MINOR }}
            type=raw,value=latest

      - name: Create and push multi-arch manifest
        working-directory: /tmp/digests
        env:
          META_JSON: ${{ steps.meta.outputs.json }}
        run: |
          # Word-splitting here is intentional: the jq output expands to
          # `-t tagA -t tagB ...` and the printf to space-separated
          # `image@sha256:<digest>` operands, each a distinct argv entry.
          # shellcheck disable=SC2046
          docker buildx imagetools create \
            $(jq -cr '.tags | map("-t " + .) | join(" ")' <<< "$META_JSON") \
            $(printf "${IMAGE}@sha256:%s " *)

      - name: Inspect manifest
        run: docker buildx imagetools inspect "${IMAGE}:${VERSION}"

  # ── Merge the webui image's per-arch digests ────────────────────────────────
  merge-webui:
    needs: [resolve, build-webui]
    if: github.event_name != 'pull_request'
    runs-on: ubuntu-24.04
    timeout-minutes: 10
    permissions:
      contents: read
      packages: write
    env:
      IMAGE: ${{ needs.resolve.outputs.image_webui }}
      VERSION: ${{ needs.resolve.outputs.version }}
      MAJOR_MINOR: ${{ needs.resolve.outputs.major_minor }}
    steps:
      - uses: actions/download-artifact@v8
        with:
          path: /tmp/digests
          pattern: webui-digests-*
          merge-multiple: true

      - uses: docker/setup-buildx-action@v4

      - uses: docker/login-action@v4.6.0
        with:
          registry: ghcr.io
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}

      - uses: docker/metadata-action@v6
        id: meta
        with:
          images: ${{ env.IMAGE }}
          tags: |
            type=raw,value=${{ env.VERSION }}
            type=raw,value=${{ env.MAJOR_MINOR }}
            type=raw,value=latest

      - name: Create and push multi-arch manifest
        working-directory: /tmp/digests
        env:
          META_JSON: ${{ steps.meta.outputs.json }}
        run: |
          # shellcheck disable=SC2046
          docker buildx imagetools create \
            $(jq -cr '.tags | map("-t " + .) | join(" ")' <<< "$META_JSON") \
            $(printf "${IMAGE}@sha256:%s " *)

      - name: Inspect manifest
        run: docker buildx imagetools inspect "${IMAGE}:${VERSION}"