ci: dispatch pr-test-extra.yml from pr-test.yml on the scheduled cron (#25320)

Co-authored-by: hnyls2002 <lsyincs@gmail.com>
Co-authored-by: Liangsheng Yin <hnyls2002@gmail.com>

.github/workflows/_pr-awareness-comment.yml

@@ -0,0 +1,124 @@
+name: PR Awareness Comment
+# Reusable workflow maintaining an "awareness block" at the bottom of the PR
+# body: extra-CI opt-in state + latest PR Test / PR Test Extra run URLs.
+# Called from pr-test.yml + pr-test-extra.yml after their check-changes job so
+# the block appears ~30s in, not after downstream stages. Each caller updates
+# only its own slot; the other slot is preserved verbatim.
+#
+# pulls.update fires `pull_request: edited`, which is NOT in the default
+# trigger types, so this does not loop back into the CI workflows.
+
+on:
+  workflow_call:
+    inputs:
+      workflow_kind:
+        description: "Which workflow is calling: 'pr-test' or 'pr-test-extra'"
+        required: true
+        type: string
+
+permissions:
+  pull-requests: write
+
+jobs:
+  update-pr-body:
+    if: github.event_name == 'pull_request'
+    runs-on: ubuntu-latest
+    # Serialize concurrent writers from pr-test + pr-test-extra: both race on
+    # get/modify/update of the PR body, and the later `pulls.update` clobbers
+    # the earlier writer's slot. Group is per-PR so unrelated PRs don't block.
+    # cancel-in-progress MUST be false — cancelling would drop the earlier update.
+    concurrency:
+      group: pr-awareness-${{ github.event.pull_request.number }}
+      cancel-in-progress: false
+    steps:
+      - name: Update awareness block in PR body
+        uses: actions/github-script@v7
+        env:
+          # Pass input via env (process.env) instead of `${{ }}` interpolation
+          # into the JS body, so future user-controlled values can't inject JS.
+          KIND: ${{ inputs.workflow_kind }}
+        with:
+          script: |
+            const kind = process.env.KIND;
+            const labels = context.payload.pull_request.labels.map(l => l.name);
+            const hasExtra = labels.includes('run-ci-extra');
+            const runUrl = `${context.serverUrl}/${context.repo.owner}/${context.repo.repo}/actions/runs/${context.runId}`;
+
+            const outerStart = '<!-- pr-awareness:start -->';
+            const outerEnd   = '<!-- pr-awareness:end -->';
+            const ptStart    = '<!-- slot:pr-test:start -->';
+            const ptEnd      = '<!-- slot:pr-test:end -->';
+            const peStart    = '<!-- slot:pr-test-extra:start -->';
+            const peEnd      = '<!-- slot:pr-test-extra:end -->';
+
+            const { data: pr } = await github.rest.pulls.get({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              pull_number: context.issue.number,
+            });
+            const body = pr.body || '';
+
+            // Extract existing block + slots so we preserve the other workflow's URL
+            const blockRe = new RegExp(`${outerStart}[\\s\\S]*?${outerEnd}`);
+            const existingBlock = (body.match(blockRe) || [''])[0];
+
+            function extractSlot(text, start, end) {
+              const re = new RegExp(`${start}([\\s\\S]*?)${end}`);
+              const m = text.match(re);
+              return m ? m[1] : null;
+            }
+            const prevPrTest  = extractSlot(existingBlock, ptStart, ptEnd);
+            const prevPrExtra = extractSlot(existingBlock, peStart, peEnd);
+
+            // Each caller updates only its own slot; the other slot is preserved.
+            // For pr-test-extra without the label: the rest of its graph is
+            // label-gate-skipped, so linking to that run would mislead — inline
+            // a warning notice in the slot instead.
+            // When kind=pr-test and hasExtra=false, force-reset the extra slot
+            // rather than carrying prevPrExtra forward — otherwise an old Run
+            // URL from a previous opt-in lingers after the label is removed,
+            // until pr-test-extra runs again and rewrites the slot.
+            const notEnabledText = ':warning: **Not enabled** — add `run-ci-extra` label to opt in.';
+            const newPrTest = (kind === 'pr-test')
+              ? `[Run #${context.runId}](${runUrl})`
+              : (prevPrTest || '_Not run yet_');
+            const newPrExtra = (kind === 'pr-test-extra')
+              ? (hasExtra
+                  ? `[Run #${context.runId}](${runUrl})`
+                  : notEnabledText)
+              : (hasExtra
+                  ? (prevPrExtra || '_Not run yet_')
+                  : notEnabledText);
+
+            const newBlock = [
+              outerStart,
+              '---',
+              '### CI Awareness',
+              '',
+              `Latest PR Test: ${ptStart}${newPrTest}${ptEnd}`,
+              `Latest PR Test (Extra): ${peStart}${newPrExtra}${peEnd}`,
+              outerEnd,
+            ].join('\n');
+
+            // Replace existing block in place, or append at body end.
+            let newBody;
+            if (blockRe.test(body)) {
+              newBody = body.replace(blockRe, newBlock);
+            } else {
+              const sep = body.length === 0
+                ? ''
+                : (body.endsWith('\n') ? '\n---\n\n' : '\n\n---\n\n');
+              newBody = `${body}${sep}${newBlock}`;
+            }
+
+            if (newBody !== body) {
+              await github.rest.pulls.update({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                pull_number: context.issue.number,
+                body: newBody,
+              });
+              core.info(`Updated PR body awareness block (caller=${kind}).`);
+            } else {
+              core.info(`PR body awareness block already up-to-date (caller=${kind}).`);
+            }

.github/workflows/_pr-test-check-changes.yml

@@ -49,6 +49,7 @@ on:
 
 jobs:
   run:
+    name: check-changes
     runs-on: ubuntu-latest
     outputs:
       # Use API-based detection for target_stage mode (filter-api), otherwise use dorny/paths-filter (filter)

.github/workflows/_pr-test-stage.yml

@@ -73,6 +73,7 @@ env:
 
 jobs:
   run:
+    name: ${{ inputs.self_name }} (${{ matrix.partition }})
     # target_stage takes precedence; otherwise default per-commit gating
     # runs on schedule / parallel-dispatch / non-failed PR with main_package
     # or sgl_kernel changes.

.github/workflows/pr-test-extra.yml

@@ -67,7 +67,7 @@ env:
   SGLANG_IS_IN_CI: true
   SGLANG_CUDA_COREDUMP: "1"
   SGLANG_JIT_DEEPGEMM_FAST_WARMUP: true
-  SKIP_STAGE_HEALTH_CHECK: ${{ inputs.skip_stage_health_check == true && 'true' || 'false' }}
+  SKIP_STAGE_HEALTH_CHECK: ${{ (inputs.skip_stage_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
@@ -105,6 +105,19 @@ jobs:
       pr_test_yml: '.github/workflows/pr-test-extra.yml'
     secrets: inherit
 
+  # always(): on a PR without `run-ci-extra` the rest of this workflow's graph
+  # is skipped, but we still want the awareness block updated so the author
+  # sees this run was a no-op.
+  call-pr-awareness:
+    needs: check-changes
+    if: always() && github.event_name == 'pull_request'
+    permissions:
+      pull-requests: write
+    uses: ./.github/workflows/_pr-awareness-comment.yml
+    with:
+      workflow_kind: pr-test-extra
+    secrets: inherit
+
   # =============================================== sgl-kernel ====================================================
   sgl-kernel-build-wheels:
     needs: check-changes

.github/workflows/pr-test.yml

@@ -35,6 +35,11 @@ on:
         required: false
         type: boolean
         default: false
+      run_all_tests:
+        description: "Run all tests (mirrors the scheduled cron's behavior — combine with test_parallel_dispatch for full sim)"
+        required: false
+        type: boolean
+        default: false
   workflow_call:
     inputs:
       git_ref:
@@ -67,7 +72,7 @@ env:
   SGLANG_IS_IN_CI: true
   SGLANG_CUDA_COREDUMP: "1"
   SGLANG_JIT_DEEPGEMM_FAST_WARMUP: true
-  SKIP_STAGE_HEALTH_CHECK: ${{ inputs.skip_stage_health_check == true && 'true' || 'false' }}
+  SKIP_STAGE_HEALTH_CHECK: ${{ (inputs.skip_stage_health_check == true || inputs.test_parallel_dispatch == true || inputs.run_all_tests == true) && 'true' || 'false' }}
   # TEMP: rebuild deepep against the new torch for torch-211-merge PR only — revert before merging to main.
   FORCE_REBUILD_DEEPEP: '1'
   # Schedule / main-branch dispatch / workflow_call from main use refs/heads/main; PR events use refs/pull/*/merge
@@ -93,6 +98,25 @@ jobs:
       force_continue_on_error: ${{ inputs.force_continue_on_error == true }}
     secrets: inherit
 
+  # Runs right after check-changes so the awareness block is injected ~30s in
+  # rather than waiting for downstream stages.
+  call-pr-awareness:
+    needs: check-changes
+    if: always() && github.event_name == 'pull_request'
+    permissions:
+      pull-requests: write
+    uses: ./.github/workflows/_pr-awareness-comment.yml
+    with:
+      workflow_kind: pr-test
+    secrets: inherit
+
+  call-pr-test-extra:
+    if: github.event_name == 'schedule' || inputs.test_parallel_dispatch == true
+    uses: ./.github/workflows/pr-test-extra.yml
+    with:
+      run_all_tests: true
+    secrets: inherit
+
   # =============================================== Wait Jobs for Sequential PR Execution ====================================================
   # These jobs poll GitHub API to wait for previous stages to complete.
   # For PR runs: wait jobs run and enforce sequential execution via polling.