test(hosted): production e2e for scan --mode hosted across every ecosystem and package manager

.github/workflows/ci.yml

@@ -4,6 +4,15 @@ on:
   push:
     branches: [main]
   pull_request:
+  workflow_dispatch:
+    inputs:
+      hosted_e2e:
+        # Underscored on purpose: `inputs.hosted-e2e` is not valid expression
+        # syntax (a hyphenated name needs `inputs['hosted-e2e']`).
+        description: 'hosted-e2e: auto (obey vars.HOSTED_E2E_DISABLED) | force | skip'
+        type: choice
+        default: auto
+        options: [auto, force, skip]
 
 permissions:
   contents: read
@@ -755,4 +764,194 @@ jobs:
         with:
           name: setup-matrix-${{ matrix.ecosystem }}
           path: report-${{ matrix.ecosystem }}.json
+
+  # ----------------------------------------------------------------------
+  # Hosted-mode production e2e — REQUIRED status check, with a kill switch.
+  #
+  # Drives `scan --mode hosted` against the REAL production endpoints
+  # (patches-api.socket.dev + patch.socket.dev) and the REAL upstream
+  # registries, using patches that are actually published on production.
+  # Nothing is mocked. Every other hosted-mode capstone in this repo
+  # (e2e_redirect_*) points at a wiremock stand-in, so this job is the only
+  # thing that would notice production drifting away from the CLI.
+  #
+  # The suite itself is `#[ignore]`-gated, so it stays OUT of the `test` and
+  # `e2e` jobs and only runs where it is explicitly asked for — here.
+  #
+  # INVARIANTS (this job is registered in branch protection as a required
+  # check named exactly `hosted-e2e`):
+  #   * NO job-level `if:` — a *skipped* required check is ambiguous to branch
+  #     protection and can wedge a PR at "Expected — waiting for status".
+  #     The kill switch gates the STEPS, never the job.
+  #   * NO `needs:` — an upstream failure would skip this job, same wedge.
+  #   * NO matrix and NO rename — the check name must stay `hosted-e2e`.
+  #   * NO `continue-on-error` — a bypass must be visible, not invisible.
+  # The job ALWAYS runs and ALWAYS reaches success or failure.
+  #
+  # ESCAPE HATCH — when production is down and this is blocking merges:
+  #   Settings -> Secrets and variables -> Actions -> Variables ->
+  #     HOSTED_E2E_DISABLED = true
+  #   then "Re-run failed jobs" on any blocked PR. `vars` is read at job-run
+  #   time, so no commit and no push is needed; the job goes green with a loud
+  #   ::warning:: and a BYPASSED banner in the job summary. DELETE the variable
+  #   to re-arm. For a one-off: Actions -> CI -> Run workflow ->
+  #   hosted_e2e = force (ignore the variable) | skip (bypass this run).
+  # ----------------------------------------------------------------------
+  hosted-e2e:
+    name: hosted-e2e            # registered in branch protection; do not rename
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    timeout-minutes: 30
+    concurrency:
+      # These are real requests against a real production service — keep it to
+      # one run per ref rather than one per push.
+      group: hosted-e2e-${{ github.ref }}
+      cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+    env:
+      HOSTED_E2E_DISABLED: ${{ vars.HOSTED_E2E_DISABLED }}
+      # The `inputs` context is empty on push/pull_request, so default to auto.
+      HOSTED_E2E_MODE: ${{ (github.event_name == 'workflow_dispatch' && inputs.hosted_e2e) || 'auto' }}
+    steps:
+      - name: Resolve the kill switch
+        id: gate
+        run: |
+          set -eu
+          run=true; reason=''
+          case "$HOSTED_E2E_MODE" in
+            force) reason='workflow_dispatch hosted_e2e=force (kill switch ignored)' ;;
+            skip)  run=false; reason='workflow_dispatch hosted_e2e=skip' ;;
+            *)     if [ "${HOSTED_E2E_DISABLED:-}" = 'true' ]; then
+                     run=false
+                     reason='repository variable HOSTED_E2E_DISABLED=true'
+                   fi ;;
+          esac
+          echo "run=$run" >> "$GITHUB_OUTPUT"
+          if [ "$run" = 'false' ]; then
+            echo "::warning title=hosted-e2e BYPASSED::$reason"
+            {
+              echo '## :warning: hosted-e2e BYPASSED — no production coverage in this run'
+              echo
+              echo "Reason: $reason"
+              echo
+              echo 'Re-arm by deleting the HOSTED_E2E_DISABLED repository variable'
+              echo '(Settings -> Secrets and variables -> Actions -> Variables).'
+            } >> "$GITHUB_STEP_SUMMARY"
+          fi
+
+      - name: Checkout
+        if: steps.gate.outputs.run == 'true'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          persist-credentials: false
+
+      - name: Install Rust
+        if: steps.gate.outputs.run == 'true'
+        run: rustup show
+
+      - name: Cache cargo
+        if: steps.gate.outputs.run == 'true'
+        uses: Swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 # v2.9.1
+        with:
+          key: hosted-e2e
+          save-if: ${{ github.ref == 'refs/heads/main' }}
+
+      - name: Setup Node.js
+        if: steps.gate.outputs.run == 'true'
+        uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
+        with:
+          # Node 24, NOT the 20.20.2 the other jobs pin. pnpm 10 imports
+          # `node:sqlite` for its store index, which does not exist before
+          # Node 22 (and is only stable in 24) — on 20 every `pnpm install`
+          # dies with ERR_UNKNOWN_BUILTIN_MODULE. This suite drives real,
+          # current package managers against production, so it tracks what
+          # users actually run rather than the pin the offline suites need.
+          node-version: '24.x'
+
+      - name: Setup npm-family package managers
+        if: steps.gate.outputs.run == 'true'
+        env:
+          # Every corepack shim invocation — including the suite's own
+          # `has_command` probes — must be non-interactive, or the probe hangs
+          # or exits non-zero and STRICT turns that into a failed leg.
+          COREPACK_ENABLE_DOWNLOAD_PROMPT: '0'
+        # corepack owns pnpm and both yarn flavors. `corepack enable` alone is
+        # not enough: it installs the shims, but `pnpm --version` still fails
+        # for a project with no `packageManager` field — exactly the pnpm
+        # fixture's shape — because the shim has no version to resolve. The
+        # `prepare … --activate` lines set that global default AND pre-download
+        # each version, so the first invocation inside a test is not also a
+        # network fetch.
+        #
+        # pnpm must NOT come from `npm install -g` here: corepack has already
+        # created its shim at the same path, and npm refuses with EEXIST. Only
+        # bun, which corepack does not manage, comes from npm.
+        run: |
+          set -eu
+          corepack enable
+          corepack prepare pnpm@10 --activate
+          corepack prepare yarn@1.22.22 --activate
+          corepack prepare yarn@4.6.0 --activate
+          npm install -g bun@1
+          node --version
+          npm --version
+          pnpm --version
+          bun --version
+
+      - name: Setup Python + uv
+        if: steps.gate.outputs.run == 'true'
+        uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5
+        with:
+          python-version: '3.12.x'
+
+      - name: Install uv
+        if: steps.gate.outputs.run == 'true'
+        run: python -m pip install --disable-pip-version-check uv && uv --version
+
+      - name: Setup Ruby
+        if: steps.gate.outputs.run == 'true'
+        uses: ruby/setup-ruby@319994f95fa847cf3fb3cd3dbe89f6dcde9f178f # v1.295.0
+        with:
+          ruby-version: '3.2.10'
+          # The gem hosted rewrite pins into the Gemfile.lock CHECKSUMS
+          # section, which `bundle lock --add-checksums` only emits on >= 2.6.
+          bundler: '2.6'
+          bundler-cache: false
+
+      - name: Setup Go
+        if: steps.gate.outputs.run == 'true'
+        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6.4.0
+        with:
+          go-version: '1.24'
+          cache: false
+
+      - name: Run hosted-mode production e2e
+        if: steps.gate.outputs.run == 'true'
+        env:
+          # A required check must never report green on an unexercised leg:
+          # STRICT turns the suite's local "toolchain missing" soft-skips into
+          # hard failures. Every toolchain it needs is installed above.
+          SOCKET_PATCH_HOSTED_E2E_STRICT: '1'
+          COREPACK_ENABLE_DOWNLOAD_PROMPT: '0'
+        run: |
+          set -u
+          # The public proxy intermittently returns 503 "Service temporarily
+          # over capacity" — that is the documented reason the older live-API
+          # suites were pulled from the PR matrix (see the `e2e` job). Retry the
+          # whole suite a couple of times before calling it a real failure, so a
+          # transient 503 does not block merges through a required check.
+          for attempt in 1 2 3; do
+            echo "::group::hosted-e2e attempt $attempt"
+            cargo test -p socket-patch-cli --test e2e_hosted_production -- \
+              --ignored --nocapture --test-threads=4
+            status=$?
+            echo "::endgroup::"
+            if [ "$status" -eq 0 ]; then
+              exit 0
+            fi
+            echo "::warning title=hosted-e2e attempt $attempt failed::retrying"
+            sleep $((attempt * 20))
+          done
+          echo "::error title=hosted-e2e::suite failed on all 3 attempts"
+          exit 1
           if-no-files-found: warn