[lint] workflow consistency linter to look at all files instead of just changed files (#165171)

clee2000 · pytorchmergebot · commit 32ac38f85d53 · 2025-10-24T21:43:18.000Z
As in title If you change only one workflow file, lintrunner (default arg, also the one in CI since it only inputs changed files) won't look at other files in the repo, but the sync-tag might come from those other files This makes it so that it looks at all workflow files so it will catch those failures Also change output line so it prints which file + which job it is different from Pros: catches errors Cons: unusual behavior (getting around what lintrunner says the linter should run on) Pull Request resolved: #165171 Approved by: https://github.com/malfet, https://github.com/izaitsevfb, https://github.com/atalman
diff --git a/tools/linter/adapters/workflow_consistency_linter.py b/tools/linter/adapters/workflow_consistency_linter.py
@@ -16,6 +16,9 @@
 from yaml import dump, load
 
 
+REPO_ROOT = Path(__file__).resolve().parent.parent.parent.parent
+
+
 if TYPE_CHECKING:
     from collections.abc import Iterable
 
@@ -59,7 +62,13 @@ def is_workflow(yaml: Any) -> bool:
     return yaml.get("jobs") is not None
 
 
-def print_lint_message(path: Path, job: dict[str, Any], sync_tag: str) -> None:
+def print_lint_message(
+    path: Path,
+    job: dict[str, Any],
+    sync_tag: str,
+    baseline_path: Path,
+    baseline_job_id: str,
+) -> None:
     job_id = next(iter(job.keys()))
     with open(path) as f:
         lines = f.readlines()
@@ -77,11 +86,30 @@ def print_lint_message(path: Path, job: dict[str, Any], sync_tag: str) -> None:
         name="workflow-inconsistency",
         original=None,
         replacement=None,
-        description=f"Job doesn't match other jobs with sync-tag: '{sync_tag}'",
+        description=f"Job doesn't match other job {baseline_job_id} in file {baseline_path} with sync-tag: '{sync_tag}'",
     )
     print(json.dumps(lint_message._asdict()), flush=True)
 
 
+def get_jobs_with_sync_tag(
+    job: dict[str, Any],
+) -> tuple[str, str, dict[str, Any]] | None:
+    sync_tag = job.get("with", {}).get("sync-tag")
+    if sync_tag is None:
+        return None
+
+    # remove the "if" field, which we allow to be different between jobs
+    # (since you might have different triggering conditions on pull vs.
+    # trunk, say.)
+    if "if" in job:
+        del job["if"]
+
+    # same is true for ['with']['test-matrix']
+    if "test-matrix" in job.get("with", {}):
+        del job["with"]["test-matrix"]
+    return (sync_tag, job_id, job)
+
+
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(
         description="workflow consistency linter.",
@@ -94,41 +122,39 @@ def print_lint_message(path: Path, job: dict[str, Any], sync_tag: str) -> None:
     )
     args = parser.parse_args()
 
-    # Go through the provided files, aggregating jobs with the same sync tag
+    # Go through all files, aggregating jobs with the same sync tag
     tag_to_jobs = defaultdict(list)
+    for path in REPO_ROOT.glob(".github/workflows/*"):
+        if not path.is_file() or path.suffix not in {".yml", ".yaml"}:
+            continue
+        workflow = load_yaml(path)
+        if not is_workflow(workflow):
+            continue
+        clean_path = path.relative_to(REPO_ROOT)
+        jobs = workflow.get("jobs", {})
+        for job_id, job in jobs.items():
+            res = get_jobs_with_sync_tag(job)
+            if res is None:
+                continue
+            sync_tag, job_id, job_dict = res
+            tag_to_jobs[sync_tag].append((clean_path, job_id, job_dict))
+
+    # Check the files passed as arguments
     for path in args.filenames:
         workflow = load_yaml(Path(path))
         jobs = workflow["jobs"]
         for job_id, job in jobs.items():
-            try:
-                sync_tag = job["with"]["sync-tag"]
-            except KeyError:
+            res = get_jobs_with_sync_tag(job)
+            if res is None:
                 continue
-
-            # remove the "if" field, which we allow to be different between jobs
-            # (since you might have different triggering conditions on pull vs.
-            # trunk, say.)
-            if "if" in job:
-                del job["if"]
-
-            # same is true for ['with']['test-matrix']
-            if "test-matrix" in job.get("with", {}):
-                del job["with"]["test-matrix"]
-
-            tag_to_jobs[sync_tag].append((path, {job_id: job}))
-
-    # For each sync tag, check that all the jobs have the same code.
-    for sync_tag, path_and_jobs in tag_to_jobs.items():
-        baseline_path, baseline_dict = path_and_jobs.pop()
-        baseline_str = dump(baseline_dict)
-
-        printed_baseline = False
-
-        for path, job_dict in path_and_jobs:
+            sync_tag, job_id, job_dict = res
             job_str = dump(job_dict)
-            if baseline_str != job_str:
-                print_lint_message(path, job_dict, sync_tag)
 
-                if not printed_baseline:
-                    print_lint_message(baseline_path, baseline_dict, sync_tag)
-                    printed_baseline = True
+            # For each sync tag, check that all the jobs have the same code.
+            for baseline_path, baseline_job_id, baseline_dict in tag_to_jobs[sync_tag]:
+                baseline_str = dump(baseline_dict)
+
+                if job_id != baseline_job_id or job_str != baseline_str:
+                    print_lint_message(
+                        path, job_dict, sync_tag, baseline_path, baseline_job_id
+                    )
