[None][infra] improve CBTS hits for formatting and VisualGen changes

jenkins/scripts/cbts/README.md

@@ -48,7 +48,7 @@ Nine rules, registered in `cbts/command/main.py::RULE_CLASSES`:
 | `WaivesRule` | `waiveonly` | `tests/integration/test_lists/waives.txt` |
 | `TestsDefRule` | `testdefonly` | `tests/**/*` (.py via AST; data files via dir walk-up) |
 | `TestListRule` | `testlistonly` | `tests/integration/test_lists/test-db/*.yml` |
-| `VisualGenRule` | `visualgenonly` | `examples/visual_gen/**`, `scripts/visualgen_eval/**`, `tensorrt_llm/_torch/visual_gen/**`, `tensorrt_llm/media/**`, `tensorrt_llm/visual_gen/**` (excl. docs; reference images such as `cat_piano.png` ARE test fixtures and stay claimed; outward-facing files force fallback) |
+| `VisualGenRule` | `visualgenonly` | VG source and the exact VG serve schema; public VG/media changes also keep full CPU and shared serving tests, package sanity, and performance sanity; missing required validation forces fallback |
 | `SpecDecRule` | `specdeconly` | `tensorrt_llm/_torch/speculative/**`, `tensorrt_llm/models/{eagle,medusa,redrafter}/**`, `examples/{eagle,medusa,redrafter,draft_target_model,ngram}/**`, `examples/llm-api/llm_speculative_decoding.py` (excl. docs; other suffixes incl. images kept as potential test fixtures) |
 | `ModelingV2Rule` | `modelingv2only` | `tensorrt_llm/_torch/_experimental/modeling_v2/**`, `cpp/tensorrt_llm/modeling_v2/**`, `cpp/cmake/modules/modeling_v2.cmake` (excl. docs); per model node for changes under `models/` |
 | `AgentFlowRule` | `agentflowonly` | `agent-flow/**` (excl. docs and out-of-scope metadata) |
@@ -65,7 +65,7 @@ See `cbts/rules/README.md` for per-rule logic.
 | `waiveonly` | `WaivesRule` fired solo: PR only edits `waives.txt`. |
 | `testdefonly` | `TestsDefRule` fired solo: PR only edits files under `tests/**/*`. |
 | `testlistonly` | `TestListRule` fired solo: PR only adds entries under `tests/integration/test_lists/test-db/*.yml`. |
-| `visualgenonly` | `VisualGenRule` fired solo: PR only touches VisualGen internal source paths (`examples/visual_gen/**`, `scripts/visualgen_eval/**`, `tensorrt_llm/_torch/visual_gen/**`; excl. docs; image fixtures like `cat_piano.png` are claimed). Narrows to blocks containing VG test entries. Outward-facing files under `tensorrt_llm/visual_gen/**` and `tensorrt_llm/media/**` (eagerly imported by `trtllm-serve`) force `null` fallback. |
+| `visualgenonly` | `VisualGenRule` fired solo. Internal VG changes select VG test entries; schema changes also retain schema validation. Public VG/media changes additionally retain the complete CPU suite, shared serving integration tests, package sanity, and performance sanity. Missing required validation forces `null` fallback. |
 | `specdeconly` | `SpecDecRule` fired solo: PR only touches speculative-decoding source paths (`tensorrt_llm/_torch/speculative/**`, `tensorrt_llm/models/{eagle,medusa,redrafter}/**`, `examples/{eagle,medusa,redrafter,draft_target_model,ngram}/**`, `examples/llm-api/llm_speculative_decoding.py`; excl. docs). Narrows to blocks containing spec-dec test entries (eagle / medusa / redrafter / ngram / draft-target-model / MTP). |
 | `modelingv2only` | `ModelingV2Rule` fired solo: PR only touches modeling_v2 code -- the subtree (`tensorrt_llm/_torch/_experimental/modeling_v2/**`, each family's native code in its `csrc/` included: one torch-op library per model family) or what the family libraries share (`cpp/tensorrt_llm/modeling_v2/**`, `cpp/cmake/modules/modeling_v2.cmake`); excl. docs, which are a fifth of the Python subtree — every catalog entry carries a contract document. A change under `models/` selects its model node's unit tests and target accuracy gates, plus whole-tree checks; shared code or an unreadable routing tree selects every modeling_v2 entry. No outward-facing fallback is needed: nothing beyond the resolver module is imported, and no family library is loaded, unless `LLM(modeling_v2=...)` is on; the CPU checks that keep this true run in every scope. The callers outside the subtree (`_torch/pyexecutor/model_loader.py`, `_torch/models/modeling_auto.py`) are left unclaimed, so touching them falls back to baseline. |
 | `agentflowonly` | `AgentFlowRule` fired solo: PR only touches `agent-flow/**` source or test files (excl. docs). Runs `CPU-AgentFlow-UnitTest`. |

jenkins/scripts/cbts/cbts/command/main.py

@@ -31,6 +31,7 @@
 
 import json
 import sys
+from collections import Counter
 from dataclasses import dataclass, field, replace
 from pathlib import Path
 from typing import Optional
@@ -64,6 +65,7 @@
 from cbts.rules.modeling_v2 import ModelingV2Rule
 from cbts.rules.openengine import OpenEngineRule
 from cbts.rules.out_of_scope import OutOfScopeRule, is_out_of_scope
+from cbts.rules.python_formatting import PythonFormattingRule, is_formatting_only_change
 from cbts.rules.spec_dec import SpecDecRule
 from cbts.rules.test_list import TestListRule
 from cbts.rules.tests_def import TestsDefRule
@@ -77,6 +79,7 @@
 
 # Classes are used for `--list-needed-diffs` (no need to construct).
 RULE_CLASSES: list[type[Rule]] = [
+    PythonFormattingRule,
     WaivesRule,
     TestsDefRule,
     TestListRule,
@@ -96,6 +99,7 @@ def build_rules(
     repo_root: Path,
 ) -> list[Rule]:
     return [
+        PythonFormattingRule(),
         WaivesRule(yaml_index, stages),
         TestsDefRule(yaml_index, stages, repo_root=repo_root),
         TestListRule(yaml_index, stages, repo_root=repo_root),
@@ -274,7 +278,9 @@ def run(self, pr: PRInputs, rules: list[Rule]) -> SelectionResult:
         code_pr = replace(
             pr,
             changed_files=[
-                path for path in pr.changed_files if path not in pr.docstring_only_files
+                path
+                for path in pr.changed_files
+                if path not in pr.docstring_only_files | pr.formatting_only_files
             ],
         )
         pairs: list[tuple[Rule, RuleResult]] = []
@@ -365,16 +371,23 @@ def _load_pr_inputs(input_json_path: Path, repo_root: Path) -> PRInputs:
     # A null diff (PR API omits the patch for binary, rename, or
     # too-large diffs) normalizes to an empty diff.
     raw_diffs = {path: d or "" for path, d in data.get("diffs", {}).items()}
+    raw_diff_counts = Counter(raw_diffs.values())
     diffs = {path: strip_noop_diff_lines(diff) for path, diff in raw_diffs.items()}
     docstring_files: set[str] = set()
     docstring_only_files: set[str] = set()
+    formatting_only_files: set[str] = set()
     for path, raw_diff in raw_diffs.items():
         if not path.endswith(".py") or not raw_diff or is_out_of_scope(path):
             continue
         try:
             source = (repo_root / path).read_text(encoding="utf-8")
         except (OSError, UnicodeDecodeError):
             continue
+        # Forge APIs repeat a renamed file's patch for its old and new paths.
+        # Content equivalence alone cannot prove that such a move has no impact.
+        if raw_diff_counts[raw_diff] == 1 and is_formatting_only_change(source, raw_diff):
+            formatting_only_files.add(path)
+            continue
         executable, changed = split_docstring_diff(source, raw_diff)
         if changed:
             docstring_files.add(path)
@@ -388,6 +401,7 @@ def _load_pr_inputs(input_json_path: Path, repo_root: Path) -> PRInputs:
         raw_diffs=raw_diffs,
         docstring_files=docstring_files,
         docstring_only_files=docstring_only_files,
+        formatting_only_files=formatting_only_files,
     )
 
 

jenkins/scripts/cbts/cbts/rules/README.md

@@ -11,17 +11,31 @@ for the overall CBTS architecture.
 
 | File | Class | Scope | Triggers on |
 |---|---|---|---|
+| `python_formatting.py` | `PythonFormattingRule` | `noop` | Python files with verified equivalent complete ASTs and unchanged comments |
 | `waives.py` | `WaivesRule` | `waiveonly` | `tests/integration/test_lists/waives.txt` |
 | `tests_def.py` | `TestsDefRule` | `testdefonly` | `tests/**/*` (any file under tests/) |
 | `test_list.py` | `TestListRule` | `testlistonly` | `tests/integration/test_lists/test-db/*.yml` |
-| `visual_gen.py` | `VisualGenRule` | `visualgenonly` | `examples/visual_gen/**`, `scripts/visualgen_eval/**`, `tensorrt_llm/_torch/visual_gen/**`, `tensorrt_llm/media/**`, `tensorrt_llm/visual_gen/**` (each excl. docs) |
+| `visual_gen.py` | `VisualGenRule` | `visualgenonly` | `examples/visual_gen/**`, `scripts/visualgen_eval/**`, `tensorrt_llm/_torch/visual_gen/**`, `tensorrt_llm/media/**`, `tensorrt_llm/visual_gen/**` (each excl. docs), and the exact VisualGen serve schema |
 | `spec_dec.py` | `SpecDecRule` | `specdeconly` | `tensorrt_llm/_torch/speculative/**`, `tensorrt_llm/models/{eagle,medusa,redrafter}/**`, `examples/{eagle,medusa,redrafter,draft_target_model,ngram}/**`, `examples/llm-api/llm_speculative_decoding.py` (each excl. docs) |
 | `modeling_v2.py` | `ModelingV2Rule` | `modelingv2only` | `tensorrt_llm/_torch/_experimental/modeling_v2/**`, `cpp/tensorrt_llm/modeling_v2/**`, `cpp/cmake/modules/modeling_v2.cmake` (excl. docs); per model node for changes under `models/` |
 | `agent_flow.py` | `AgentFlowRule` | `agentflowonly` | `agent-flow/**` (excl. docs and out-of-scope metadata) → the single `CPU-AgentFlow-UnitTest` stage; not test-db-driven |
 | `openengine.py` | `OpenEngineRule` | `openengineonly` | `tensorrt_llm/grpc/openengine/**` (excl. docs) → the `l0_cpu` block containing `unittest/grpc/openengine/` |
 | `docs.py` | `DocsRule` | `docsonly` | Outside noop paths: `docs/**` and verified Python docstring edits → `CPU-Build_Docs` plus complete `l0_cpu`; other `**/*.md` / `**/*.rst` → `CPU-Build_Docs` only |
 | `out_of_scope.py` | `OutOfScopeRule` | `noop` | Reviewed metadata paths, root issue-template and developer-assistant folders, `.gitignore` at any depth, conventional clang filenames under `cpp/`, verified static-check-only `pyproject.toml` edits, QA/dev lists, timing caches, microbenchmarks, and CBTS infrastructure |
 
+## PythonFormattingRule
+
+Claims Python files only after the raw patch matches the checked-out source
+and both complete source images have identical ASTs, including docstrings and
+type comments. Comment text must also match, preserving encoding and lint
+directives. Missing or mismatched patches, syntax errors, empty pre-images,
+and explicit rename/mode changes retain their normal impact. Repeated patches
+for old/new rename paths are not classified as formatting.
+
+Verified files are excluded from source rules and contribute `scope=noop`;
+the entry point still retains the always-run CPU stages. A mixed PR continues
+through the normal rules for every remaining changed file.
+
 ## DocsRule
 
 Claims every file under `docs/` plus Markdown and reStructuredText files
@@ -203,8 +217,10 @@ Outcomes:
 
 Path-only rule. Claims source files under `examples/visual_gen/`,
 `scripts/visualgen_eval/`, `tensorrt_llm/_torch/visual_gen/`,
-`tensorrt_llm/media/`, and `tensorrt_llm/visual_gen/` (excluding `.md`,
-which `OutOfScopeRule` claims as noop). fused-DiT paths stay unclaimed:
+`tensorrt_llm/media/`, and `tensorrt_llm/visual_gen/` (excluding documentation,
+which `DocsRule` handles). It also claims the exact generated schema
+`tensorrt_llm/schemas/trtllm-serve-visual-gen-config.schema.json`.
+Fused-DiT paths stay unclaimed:
 those diffs also touch `inplace_info()` in
 `tensorrt_llm/_torch/compilation/utils.py`, which no VG rule claims.
 Image suffixes are intentionally NOT excluded: VG ships reference images
@@ -222,16 +238,25 @@ live in `backend: pytorch` and `backend: tensorrt` blocks. A block
 
 For each matched block, `block_filters` keeps only the VG entries.
 Non-VG siblings in the same block stay governed by other rules.
-
-Outward-facing fallback: VisualGen public symbols and
-`tensorrt_llm.media.*` are imported eagerly by non-VG startup paths such
-as `commands/serve.py`, `commands/utils.py`, and
-`serve/openai_server.py`. Both package prefixes
-(`tensorrt_llm/visual_gen/`, `tensorrt_llm/media/`) are listed in
-`_VG_OUTWARD_PREFIXES`; touching any non-doc file under them claims the
-changed files but emits `scope=None` so Selector falls back to baseline.
-This protects trtllm-serve / trtllm-bench startup paths from VG
-signature drift slipping through pre-merge.
+Schema changes additionally keep the complete
+`unittest/llmapi/test_trtllm_serve_config_schemas.py` validation entry (or
+its enclosing test-directory entry). Missing validation stages force fallback.
+Other generated schemas remain unclaimed.
+
+Shared serving validation: changes under `tensorrt_llm/visual_gen/` or
+`tensorrt_llm/media/` also keep every entry in `l0_cpu`, plus shared test
+families under `test_e2e.py`, `examples/serve`, `disaggregated`, `llmapi`,
+`unittest/llmapi`, and `unittest/disaggregated`. These cover public config,
+protocol/media imports, normal and disaggregated serving, multimodal requests,
+and online benchmarks. PackageSanityCheck and PerfSanity remain enabled.
+
+The CPU suite must contain runnable API stability, serve-schema, lazy-server
+import, and lazy-model-zoo validation. The serving chat/completions examples and
+config-override precedence test must also resolve to stages scheduled in the trigger mode;
+OnDemand stages and optional
+pre-merge multi-GPU stages cannot satisfy this requirement. Missing or filtered
+required validation keeps `scope=None`. Public changes therefore get a wider
+test set than internal VG changes without requiring all unrelated GPU tests.
 
 Outcomes:
 
@@ -242,8 +267,9 @@ Outcomes:
   sanity off (VG changes don't affect wheel sanity); perfsanity on iff
   a matched block lives in `l0_perf` or `*perf_sanity*`.
 - VG source touched, any outward-facing path under
-  `tensorrt_llm/visual_gen/` or `tensorrt_llm/media/` → `scope=None`
-  (fallback).
+  `tensorrt_llm/visual_gen/` or `tensorrt_llm/media/` → `scope=visualgenonly`
+  with shared serving, full CPU, package sanity, and performance sanity.
+  Missing required shared validation → `scope=None` (fallback).
 - VG source touched but no VG block found anywhere (defensive) →
   `scope=None` (fallback).
 

jenkins/scripts/cbts/cbts/rules/base.py

@@ -42,7 +42,7 @@ def format_reason(reason: object) -> str:
 
 @dataclass
 class PRInputs:
-    """PR data, with executable selection diffs and verified docstring facts.
+    """PR data, with executable diffs and verified docstring/formatting facts.
 
     ``raw_diffs`` preserves source reconstruction input. ``post_merge``
     reflects /bot run --post-merge.
@@ -55,6 +55,7 @@ class PRInputs:
     # Documentation contributions can coexist with executable edits in one file.
     docstring_files: set[str] = field(default_factory=set)
     docstring_only_files: set[str] = field(default_factory=set)
+    formatting_only_files: set[str] = field(default_factory=set)
 
 
 @dataclass

jenkins/scripts/cbts/cbts/rules/python_formatting.py

@@ -0,0 +1,87 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Recognize verified Python formatting changes without a coverage dependency."""
+
+from __future__ import annotations
+
+import ast
+import io
+import tokenize
+
+from ._helpers import reconstruct_diff_pre_image
+from .base import PRInputs, Rule, RuleResult
+
+
+def is_formatting_only_change(post_source: str, diff: str) -> bool:
+    """Compare complete syntax trees and comments after verifying the raw patch.
+
+    Keep docstrings, type comments, and comment directives in the comparison.
+    Missing source, malformed patches, and new or removed files remain unknown.
+    """
+    if any(
+        line.startswith(
+            (
+                "new file mode ",
+                "deleted file mode ",
+                "old mode ",
+                "new mode ",
+                "rename from ",
+                "rename to ",
+                "copy from ",
+                "copy to ",
+            )
+        )
+        for line in diff.splitlines()
+    ):
+        return False
+    pre_source = reconstruct_diff_pre_image(post_source, diff)
+    if pre_source is None:
+        return False
+    try:
+        before = ast.parse(pre_source, type_comments=True)
+        after = ast.parse(post_source, type_comments=True)
+        comments = [
+            [
+                token.string
+                for token in tokenize.generate_tokens(io.StringIO(source).readline)
+                if token.type == tokenize.COMMENT
+            ]
+            for source in (pre_source, post_source)
+        ]
+    except (SyntaxError, ValueError, tokenize.TokenError):
+        return False
+    return (
+        bool(before.body)
+        and ast.dump(before, include_attributes=False) == ast.dump(after, include_attributes=False)
+        and comments[0] == comments[1]
+    )
+
+
+class PythonFormattingRule(Rule):
+    name = "pythonformatting"
+    needs_diff_for = ("**/*.py",)
+    handles_docstrings = True
+
+    def apply(self, pr: PRInputs) -> RuleResult | None:
+        if not pr.formatting_only_files:
+            return None
+        return RuleResult(
+            handled_files=set(pr.formatting_only_files),
+            affected_stages=set(),
+            scope="noop",
+            reason=f"pythonformatting: {len(pr.formatting_only_files)} verified formatting-only file(s)",
+            sanity_relevant=False,
+            perfsanity_relevant=False,
+        )