name: PR Test Extra
# Label-gated extra CI workflow opted into per-PR via labels.
#
# Adds runtime to a PR only when the author asks for it: the PR must carry
# BOTH `run-ci` (basic-CI prerequisite) and `run-ci-extra` (explicit opt-in
# to this workflow). The label check happens at runtime in pr-gate.yml via
# live `gh pr view --json labels`, so reruns after adding the labels via
# slash command pick up the new label set (workflow-level `if` checks would
# read the frozen event payload, which never updates on rerun). The same
# job graph runs unconditionally on workflow_dispatch / workflow_call so it
# can be triggered manually or chained from another workflow.
#
# Stages: extra-a (1-/2-gpu) and extra-b (4-/8-gpu) caller stubs reuse
# `_pr-test-stage.yml` and `_pr-test-check-changes.yml` from pr-test.yml.

on:
  pull_request:
    # `labeled` lets the workflow re-fire when `run-ci-extra` (or `run-ci`)
    # is added after the latest push — otherwise GHA leaves the stale
    # skipped run in place and there's no way to enable extra without
    # another push. See check-changes.if below for the matching guard.
    types: [opened, synchronize, reopened, labeled]
  workflow_dispatch:
    inputs:
      force_continue_on_error:
        description: "Force continue-on-error (test scheduled CI behavior)"
        required: false
        type: boolean
        default: false
      test_parallel_dispatch:
        description: "Test parallel dispatch behavior (simulates scheduled run)"
        required: false
        type: boolean
        default: false
      run_all_tests:
        description: "Run all tests — bypasses paths-filter (needed when dispatching against long-lived branches whose diff against main can't resolve a merge-base in a shallow clone)"
        required: false
        type: boolean
        default: false
  workflow_call:
    inputs:
      git_ref:
        description: 'Git ref (branch, tag, or SHA) to test. If not provided, uses the default branch.'
        required: false
        type: string
        default: ''
      run_all_tests:
        description: "Run all tests (for releasing or testing purpose)"
        required: false
        type: boolean
        default: false
      skip_pr_test_health_check:
        description: "Skip PR test health check fast-fail (e.g. for release branch cuts)"
        required: false
        type: boolean
        default: false

concurrency:
  group: pr-test-extra-${{ github.event_name }}-${{ github.head_ref || github.ref_name || 'default' }}-${{ inputs.git_ref || 'all' }}
  cancel-in-progress: ${{ github.event_name != 'workflow_call' }}

env:
  SGLANG_IS_IN_CI: true
  SGLANG_CUDA_COREDUMP: "1"
  SGLANG_JIT_DEEPGEMM_FAST_WARMUP: true
  SKIP_PR_TEST_HEALTH_CHECK: ${{ (inputs.skip_pr_test_health_check == true || inputs.run_all_tests == true) && 'true' || 'false' }}
  FORCE_REBUILD_DEEPEP: '1'
  PR_TEST_BYPASS_MAINTENANCE_ON_MAIN: ${{ github.ref == 'refs/heads/main' && 'true' || 'false' }}
  USE_VENV: false

permissions:
  actions: write
  contents: read
  issues: read
  pull-requests: read

jobs:
  # =============================================== check changes ====================================================
  # The actual `run-ci` + `run-ci-extra` label gate now lives in the
  # `call-gate` job below, which calls pr-gate.yml. pr-gate.yml live-fetches
  # the PR's current labels at runtime, so a rerun (e.g. via
  # /tag-and-rerun-ci extra, which adds labels via GITHUB_TOKEN — these
  # don't cascade-trigger labeled events) picks up the new label set and
  # passes. The previous workflow-level `if` checked the frozen event
  # payload, which never updates on rerun and made the slash-add-label flow
  # impossible to recover.
  #
  # We only keep one filter here: for `labeled` events, only proceed if the
  # just-added label is one of the gating labels. This prevents every
  # unrelated label addition from dispatching a full CI run.
  check-changes:
    if: |
      github.event_name != 'pull_request' ||
      github.event.action != 'labeled' ||
      github.event.label.name == 'run-ci' ||
      github.event.label.name == 'run-ci-extra'
    uses: ./.github/workflows/_pr-test-check-changes.yml
    with:
      git_ref: ${{ inputs.git_ref || '' }}
      run_all_tests: ${{ inputs.run_all_tests == true }}
      force_continue_on_error: ${{ inputs.force_continue_on_error == true }}
      pr_test_yml: '.github/workflows/pr-test-extra.yml'
    secrets: inherit

  # =============================================== PR Gate ====================================================
  # Runtime live-fetch gate. Mirrors pr-test.yml's call-gate, but additionally
  # requires `run-ci-extra`. Failure here cascades down to every test job
  # via `needs`, so the workflow ends with one red gate job (~30s) plus a
  # row of skipped test jobs instead of consuming a runner.
  call-gate:
    needs: check-changes
    if: github.event_name != 'schedule' && needs.check-changes.result == 'success'
    uses: ./.github/workflows/pr-gate.yml
    with:
      require-run-ci: true
      require-run-ci-extra: true
    secrets: inherit

  # =============================================== sgl-kernel ====================================================
  sgl-kernel-build-wheels:
    needs: [check-changes, call-gate]
    if: |
      needs.check-changes.result == 'success' &&
      (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') &&
      needs.check-changes.outputs.sgl_kernel == 'true'
    uses: ./.github/workflows/_pr-test-sgl-kernel-build.yml
    with:
      runs_on: x64-kernel-build-node
      job_display_name: Build Wheel
      git_ref: ${{ inputs.git_ref || '' }}
      skip_pr_test_health_check: ${{ inputs.skip_pr_test_health_check == true }}
    secrets: inherit

  # =============================================== extra-a (1-/2-gpu) ===============================================
  extra-a-test-1-gpu-small:
    needs: [check-changes, call-gate, sgl-kernel-build-wheels]
    if: ${{ !failure() && !cancelled() && needs.check-changes.result == 'success' && (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') }}
    uses: ./.github/workflows/_pr-test-stage.yml
    with:
      self_name: extra-a-test-1-gpu-small
      runner_config: 1-gpu-small
      check_changes: ${{ toJson(needs.check-changes.outputs) }}
      caller_inputs: ${{ toJson(inputs) }}
      partitions: ${{ needs.check-changes.outputs.partitions }}
      run_timeout_minutes: '60'
    secrets: inherit

  extra-a-test-1-gpu-large:
    needs: [check-changes, call-gate, sgl-kernel-build-wheels]
    if: ${{ !failure() && !cancelled() && needs.check-changes.result == 'success' && (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') }}
    uses: ./.github/workflows/_pr-test-stage.yml
    with:
      self_name: extra-a-test-1-gpu-large
      runner_config: 1-gpu-large
      check_changes: ${{ toJson(needs.check-changes.outputs) }}
      caller_inputs: ${{ toJson(inputs) }}
      partitions: ${{ needs.check-changes.outputs.partitions }}
      run_timeout_minutes: '60'
      timeout_per_file: '1800'
    secrets: inherit

  extra-a-test-2-gpu-large:
    needs: [check-changes, call-gate, sgl-kernel-build-wheels]
    if: ${{ !failure() && !cancelled() && needs.check-changes.result == 'success' && (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') }}
    uses: ./.github/workflows/_pr-test-stage.yml
    with:
      self_name: extra-a-test-2-gpu-large
      runner_config: 2-gpu-large
      check_changes: ${{ toJson(needs.check-changes.outputs) }}
      caller_inputs: ${{ toJson(inputs) }}
      partitions: ${{ needs.check-changes.outputs.partitions }}
      run_timeout_minutes: '60'
    secrets: inherit

  # =============================================== extra-b (4-/8-gpu) ===============================================
  extra-b-test-4-gpu-h100:
    needs: [check-changes, call-gate, sgl-kernel-build-wheels]
    if: ${{ !failure() && !cancelled() && needs.check-changes.result == 'success' && (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') }}
    uses: ./.github/workflows/_pr-test-stage.yml
    with:
      self_name: extra-b-test-4-gpu-h100
      runner_config: 4-gpu-h100
      check_changes: ${{ toJson(needs.check-changes.outputs) }}
      caller_inputs: ${{ toJson(inputs) }}
      partitions: ${{ needs.check-changes.outputs.partitions }}
      run_timeout_minutes: '60'
    secrets: inherit

  extra-b-test-4-gpu-b200:
    needs: [check-changes, call-gate, sgl-kernel-build-wheels]
    if: ${{ !failure() && !cancelled() && needs.check-changes.result == 'success' && (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') }}
    uses: ./.github/workflows/_pr-test-stage.yml
    with:
      self_name: extra-b-test-4-gpu-b200
      runner_config: 4-gpu-b200
      check_changes: ${{ toJson(needs.check-changes.outputs) }}
      caller_inputs: ${{ toJson(inputs) }}
      partitions: ${{ needs.check-changes.outputs.partitions }}
      run_timeout_minutes: '60'
      timeout_per_file: '1800'
    secrets: inherit

  extra-b-test-8-gpu-h200:
    needs: [check-changes, call-gate, sgl-kernel-build-wheels]
    if: ${{ !failure() && !cancelled() && needs.check-changes.result == 'success' && (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') }}
    uses: ./.github/workflows/_pr-test-stage.yml
    with:
      self_name: extra-b-test-8-gpu-h200
      runner_config: 8-gpu-h200
      check_changes: ${{ toJson(needs.check-changes.outputs) }}
      caller_inputs: ${{ toJson(inputs) }}
      partitions: ${{ needs.check-changes.outputs.partitions }}
      run_timeout_minutes: '60'
    secrets: inherit

  extra-b-test-deepep-4-gpu-h100:
    needs: [check-changes, call-gate, sgl-kernel-build-wheels]
    if: ${{ !failure() && !cancelled() && needs.check-changes.result == 'success' && (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') }}
    uses: ./.github/workflows/_pr-test-stage.yml
    with:
      self_name: extra-b-test-deepep-4-gpu-h100
      runner_config: deepep-4-gpu-h100
      check_changes: ${{ toJson(needs.check-changes.outputs) }}
      caller_inputs: ${{ toJson(inputs) }}
      partitions: ${{ needs.check-changes.outputs.partitions }}
      run_timeout_minutes: '60'
    secrets: inherit

  extra-b-test-deepep-4-gpu-b200:
    needs: [check-changes, call-gate, sgl-kernel-build-wheels]
    if: ${{ !failure() && !cancelled() && needs.check-changes.result == 'success' && (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') }}
    uses: ./.github/workflows/_pr-test-stage.yml
    with:
      self_name: extra-b-test-deepep-4-gpu-b200
      runner_config: deepep-4-gpu-b200
      check_changes: ${{ toJson(needs.check-changes.outputs) }}
      caller_inputs: ${{ toJson(inputs) }}
      partitions: ${{ needs.check-changes.outputs.partitions }}
      run_timeout_minutes: '60'
      timeout_per_file: '1800'
    secrets: inherit

  extra-b-test-deepep-8-gpu-h200:
    needs: [check-changes, call-gate, sgl-kernel-build-wheels]
    if: ${{ !failure() && !cancelled() && needs.check-changes.result == 'success' && (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped') }}
    uses: ./.github/workflows/_pr-test-stage.yml
    with:
      self_name: extra-b-test-deepep-8-gpu-h200
      runner_config: deepep-8-gpu-h200
      check_changes: ${{ toJson(needs.check-changes.outputs) }}
      caller_inputs: ${{ toJson(inputs) }}
      partitions: ${{ needs.check-changes.outputs.partitions }}
      run_timeout_minutes: '60'
    secrets: inherit

  # =============================================== aggregator ====================================================
  # Mirrors pr-test.yml's `pr-test-finish` so notify-pr-states below only
  # depends on one job rather than re-listing every stage. Fails if any
  # dependent failed/cancelled.
  pr-test-extra-finish:
    needs:
      [
        check-changes,
        call-gate,
        sgl-kernel-build-wheels,
        extra-a-test-1-gpu-small,
        extra-a-test-1-gpu-large,
        extra-a-test-2-gpu-large,
        extra-b-test-4-gpu-h100,
        extra-b-test-4-gpu-b200,
        extra-b-test-8-gpu-h200,
        extra-b-test-deepep-4-gpu-h100,
        extra-b-test-deepep-4-gpu-b200,
        extra-b-test-deepep-8-gpu-h200,
      ]
    if: always()
    runs-on: ubuntu-latest
    steps:
      - name: Check all dependent job statuses
        run: |
          json_needs='${{ toJson(needs) }}'
          job_names=$(echo "$json_needs" | jq -r 'keys_unsorted[]')
          for job in $job_names; do
            result=$(echo "$json_needs" | jq -r --arg j "$job" '.[$j].result')
            echo "$job: $result"
            if [[ "$result" == "failure" || "$result" == "cancelled" ]]; then
              echo "The above jobs failed."
              exit 1
            fi
          done
          echo "All jobs completed successfully"
          exit 0

  # =============================================== notify pr-states ====================================================
  # Dispatches pr-states.yml after every attempt (initial + every rerun).
  # workflow_run.completed only fires on the initial completion (one
  # workflow_run record across attempts), so we explicitly dispatch —
  # workflow_dispatch is the only event GITHUB_TOKEN can cascade-create.
  #
  # Fork PRs are excluded: their pull_request GITHUB_TOKEN is forced
  # read-only regardless of declared `actions: write`, so dispatch 403s.
  # Fork PRs get pr-states updates via the workflow_run subscription
  # there (initial completion) and pull_request_target (push / label).
  notify-pr-states:
    needs: [pr-test-extra-finish]
    if: |
      always() &&
      github.event_name == 'pull_request' &&
      github.event.pull_request.head.repo.full_name == github.repository
    runs-on: ubuntu-latest
    steps:
      - name: Dispatch pr-states refresh
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          PR_NUMBER: ${{ github.event.pull_request.number }}
        run: |
          gh workflow run pr-states.yml \
            --repo ${{ github.repository }} \
            --ref main \
            -f pr_number="$PR_NUMBER"