[CI] Enforce modern stage=/runner_config= form for dispatchable test suites (#28108)

.pre-commit-config.yaml

@@ -93,7 +93,7 @@ repos:
         files: ^\.github/workflows/.*\.yml$
         pass_filenames: false
       - id: check-registered-tests
-        name: check registered tests have CI registry
+        name: validate registered test CI registries
         entry: python3 scripts/ci/check_registered_tests.py
         language: system
         files: ^test/registered/.*\.py$

scripts/ci/check_registered_tests.py

@@ -1,7 +1,16 @@
 #!/usr/bin/env python3
 """
-Pre-commit hook: validate that all Python test files under test/registered/
-contain a CI registry call (register_cuda_ci, register_amd_ci, etc.).
+Pre-commit hook: validate CI registry calls under test/registered/.
+
+1. Every test file must contain a CI registry call (register_cuda_ci,
+   register_amd_ci, etc.).
+2. A CUDA test must not register a `{stage}-test-{runner_config}`-shaped suite
+   via the legacy single-string `suite=`: that form is not dispatchable via
+   /rerun-test. Use the modern `stage=`/`runner_config=` form instead -- it
+   resolves to the identical suite (CIRegistry.effective_suite is
+   f"{stage}-test-{runner_config}"), so same stage and same runner, just
+   /rerun-test-able. Legacy `suite=` stays valid for nightly/stress/weekly +
+   AMD/CPU/NPU suites whose names don't follow that shape.
 
 Reuses ut_parse_one_file() from ci_register.py (AST-based parsing)
 to match the same logic used by run_suite.py's collect_tests().
@@ -10,8 +19,14 @@ to match the same logic used by run_suite.py's collect_tests().
 import glob
 import importlib.util
 import os
+import re
 import sys
 
+# Suite names of the form `{stage}-test-{runner_config}` are exactly what the
+# modern stage=/runner_config= form produces, so a legacy suite= carrying this
+# shape is always expressible (and should be expressed) the modern way.
+_MODERN_SHAPE = re.compile(r"^(.+)-test-(.+)$")
+
 
 def main() -> int:
     # Import ci_register directly to avoid pulling in all of sglang
@@ -21,6 +36,7 @@ def main() -> int:
     )
     ci_register = importlib.util.module_from_spec(spec)
     spec.loader.exec_module(ci_register)
+    cuda = ci_register.HWBackend.CUDA
 
     # Same filter as run_suite.py: skip conftest.py, __init__.py, and utils.py
     files = sorted(
@@ -31,25 +47,55 @@ def main() -> int:
     if not files:
         return 0
 
-    errors = []
+    missing = []
+    legacy_shape = []  # (file, suite, stage, runner_config)
     for f in files:
         try:
             registries, _has_main_entry = ci_register.ut_parse_one_file(f)
-            if len(registries) == 0:
-                errors.append(f)
         except Exception:
             # Skip files that can't be parsed (syntax errors, etc.)
-            pass
+            continue
+        if len(registries) == 0:
+            missing.append(f)
+            continue
+        for r in registries:
+            # Pure legacy form on a CUDA registry: suite set, stage/runner unset.
+            if not (
+                r.backend == cuda
+                and r.suite is not None
+                and r.stage is None
+                and r.runner_config is None
+            ):
+                continue
+            m = _MODERN_SHAPE.match(r.suite)
+            if m:
+                legacy_shape.append((f, r.suite, m.group(1), m.group(2)))
 
-    if errors:
+    exit_code = 0
+    if missing:
         print("ERROR: Files in test/registered/ missing CI registry call:")
         print("  Move manual-only tests to test/manual/.\n")
-        for f in errors:
+        for f in missing:
             print(f"  {f}")
         print()
-        return 1
+        exit_code = 1
+    if legacy_shape:
+        print(
+            "ERROR: CUDA test(s) register a `{stage}-test-{runner_config}`-shaped "
+            'suite via the legacy `suite="..."` form, which is not dispatchable '
+            "via /rerun-test. Switch to the modern `stage=`/`runner_config=` form "
+            "(same stage, same runner):\n"
+        )
+        for f, suite, stage, runner_config in legacy_shape:
+            print(
+                f"  {f}\n"
+                f'    suite="{suite}"'
+                f'  ->  stage="{stage}", runner_config="{runner_config}"'
+            )
+        print()
+        exit_code = 1
 
-    return 0
+    return exit_code
 
 
 if __name__ == "__main__":

test/registered/tokenizer/test_multi_detokenizer.py

@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
     write_github_step_summary,
 )
 
-register_cuda_ci(est_time=211, suite="base-b-test-1-gpu-large")
+register_cuda_ci(est_time=211, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=345, suite="stage-b-test-1-gpu-small-amd")
 
 

test/registered/unit/spec/test_eagle_worker_v2_topk1_fastpath.py

@@ -19,7 +19,7 @@ from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=20, suite="base-b-test-1-gpu-small")
+register_cuda_ci(est_time=20, stage="base-b", runner_config="1-gpu-small")
 
 DEVICE = get_device()
 

test/registered/vlm/test_token_id_retokenize_e2e.py

@@ -35,7 +35,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=300, suite="base-b-test-1-gpu-large")
+register_cuda_ci(est_time=300, stage="base-b", runner_config="1-gpu-large")
 
 
 def _data_uri():