Merge branch 'main' into opt/79690

Author: nimit25

.ci/all_requirements.txt

@@ -0,0 +1,85 @@
+# Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+# See https://llvm.org/LICENSE.txt for license information.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+"""Caches .lit_test_times.txt files between premerge invocations.
+
+.lit_test_times.txt files are used by lit to order tests to best take advantage
+of parallelism. Having them around and up to date can result in a ~15%
+improvement in test times. This script downloading cached test time files and
+uploading new versions to the GCS buckets used for caching.
+"""
+
+import sys
+import os
+import logging
+import multiprocessing.pool
+import pathlib
+import platform
+import glob
+
+from google.cloud import storage
+
+GCS_PARALLELISM = 100
+
+
+def _get_blob_prefix():
+    return f"lit_timing_{platform.system().lower()}"
+
+
+def _maybe_upload_timing_file(bucket, timing_file_path):
+    if os.path.exists(timing_file_path):
+        timing_file_blob = bucket.blob(_get_blob_prefix() + "/" + timing_file_path)
+        timing_file_blob.upload_from_filename(timing_file_path)
+
+
+def upload_timing_files(storage_client, bucket_name: str):
+    bucket = storage_client.bucket(bucket_name)
+    with multiprocessing.pool.ThreadPool(GCS_PARALLELISM) as thread_pool:
+        futures = []
+        for timing_file_path in glob.glob("**/.lit_test_times.txt", recursive=True):
+            futures.append(
+                thread_pool.apply_async(
+                    _maybe_upload_timing_file, (bucket, timing_file_path)
+                )
+            )
+        for future in futures:
+            future.get()
+    print("Done uploading")
+
+
+def _maybe_download_timing_file(blob):
+    file_name = blob.name.removeprefix(_get_blob_prefix() + "/")
+    pathlib.Path(os.path.dirname(file_name)).mkdir(parents=True, exist_ok=True)
+    blob.download_to_filename(file_name)
+
+
+def download_timing_files(storage_client, bucket_name: str):
+    bucket = storage_client.bucket(bucket_name)
+    blobs = bucket.list_blobs(prefix=_get_blob_prefix())
+    with multiprocessing.pool.ThreadPool(GCS_PARALLELISM) as thread_pool:
+        futures = []
+        for timing_file_blob in blobs:
+            futures.append(
+                thread_pool.apply_async(
+                    _maybe_download_timing_file, (timing_file_blob,)
+                )
+            )
+        for future in futures:
+            future.get()
+    print("Done downloading")
+
+
+if __name__ == "__main__":
+    if len(sys.argv) != 2:
+        logging.fatal("Expected usage is cache_lit_timing_files.py <upload/download>")
+        sys.exit(1)
+    action = sys.argv[1]
+    storage_client = storage.Client()
+    bucket_name = os.environ["CACHE_GCS_BUCKET"]
+    if action == "download":
+        download_timing_files(storage_client, bucket_name)
+    elif action == "upload":
+        upload_timing_files(storage_client, bucket_name)
+    else:
+        logging.fatal("Expected usage is cache_lit_timing_files.py <upload/download>")
+        sys.exit(1)

.ci/cache_lit_timing_files.py

@@ -27,6 +27,13 @@ def _parse_ninja_log(ninja_log: list[str]) -> list[tuple[str, str]]:
             # We hit the end of the log without finding a build failure, go to
             # the next log.
             return failures
+        # If we are doing a build with LLVM_ENABLE_RUNTIMES, we can have nested
+        # ninja invocations. The sub-ninja will print that a subcommand failed,
+        # and then the outer ninja will list the command that failed. We should
+        # ignore the outer failure.
+        if ninja_log[index - 1].startswith("ninja: build stopped:"):
+            index += 1
+            continue
         # We are trying to parse cases like the following:
         #
         # [4/5] test/4.stamp

.ci/generate_test_report_lib.py

@@ -126,6 +126,39 @@ def test_ninja_log_multiple_failures(self):
             ),
         )
 
+    # Test that we can correctly handle the runtimes build. the LLVM runtimes
+    # build will involve ninja invoking more ninja processes within the
+    # runtimes directory. This means that we see two failures for a failure in
+    # the runtimes build: one from the inner ninja containing the actual action
+    # that failed, and one for the sub ninja invocation that failed.
+    def test_ninja_log_runtimes_failure(self):
+        failures = generate_test_report_lib.find_failure_in_ninja_logs(
+            [
+                [
+                    "[1/5] test/1.stamp",
+                    "[2/5] test/2.stamp",
+                    "FAILED: touch test/2.stamp",
+                    "Wow! This system is really broken!",
+                    "ninja: build stopped: subcommand failed.",
+                    "FAILED: running check-runtime failed.",
+                    "<some random command>",
+                    "ninja: build stopped: subcommand failed.",
+                ]
+            ]
+        )
+        self.assertEqual(len(failures), 1)
+        self.assertEqual(
+            failures[0],
+            (
+                "test/2.stamp",
+                dedent(
+                    """\
+                    FAILED: touch test/2.stamp
+                    Wow! This system is really broken!"""
+                ),
+            ),
+        )
+
     def test_title_only(self):
         self.assertEqual(
             generate_test_report_lib.generate_report("Foo", 0, [], []),

.ci/generate_test_report_lib_test.py

@@ -70,6 +70,7 @@
 # by trial and error).
 GRAFANA_METRIC_MAX_AGE_MN = 120
 
+
 @dataclass
 class JobMetrics:
     job_name: str
@@ -243,6 +244,7 @@ def clean_up_libcxx_job_name(old_name: str) -> str:
     new_name = stage + "_" + remainder
     return new_name
 
+
 def github_get_metrics(
     github_repo: github.Repository, last_workflows_seen_as_completed: set[int]
 ) -> tuple[list[JobMetrics], int]:

.ci/metrics/metrics.py

@@ -409,5 +409,6 @@ def test_clean_up_libcxx_job_name(self):
         out_name4 = metrics.clean_up_libcxx_job_name(bad_name)
         self.assertEqual(out_name4, bad_name)
 
+
 if __name__ == "__main__":
     unittest.main()

.ci/metrics/metrics_test.py

@@ -1 +1,2 @@
 junitparser==3.2.0
+google-cloud-storage==3.3.0

.ci/requirements.txt

@@ -2,7 +2,7 @@ FROM docker.io/library/ubuntu:24.04 as base
 ENV LLVM_SYSROOT=/opt/llvm
 
 FROM base as stage1-toolchain
-ENV LLVM_VERSION=20.1.8
+ENV LLVM_VERSION=21.1.0
 
 RUN apt-get update && \
     apt-get install -y \

.github/workflows/containers/github-action-ci/Dockerfile

@@ -52,10 +52,13 @@ jobs:
           echo "Formatting files:"
           echo "$CHANGED_FILES"
 
+      # The clang format version should always be upgraded to the first version
+      # of a release cycle (x.1.0) or the last version of a release cycle, or
+      # if there have been relevant clang-format backports.
       - name: Install clang-format
         uses: aminya/setup-cpp@17c11551771948abc5752bbf3183482567c7caf0 # v1.1.1
         with:
-          clangformat: 20.1.8
+          clangformat: 21.1.0
 
       - name: Setup Python env
         uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0

.github/workflows/pr-code-format.yml

@@ -110,6 +110,25 @@ AST_MATCHER_P(UserDefinedLiteral, hasLiteral,
   return false;
 }
 
+AST_MATCHER_P(CXXMethodDecl, hasCanonicalDecl,
+              clang::ast_matchers::internal::Matcher<CXXMethodDecl>,
+              InnerMatcher) {
+  return InnerMatcher.matches(*Node.getCanonicalDecl(), Finder, Builder);
+}
+
+AST_POLYMORPHIC_MATCHER_P(
+    matchMemberName,
+    AST_POLYMORPHIC_SUPPORTED_TYPES(MemberExpr, CXXDependentScopeMemberExpr),
+    std::string, MemberName) {
+  if (const auto *E = dyn_cast<MemberExpr>(&Node))
+    return E->getMemberDecl()->getName() == MemberName;
+
+  if (const auto *E = dyn_cast<CXXDependentScopeMemberExpr>(&Node))
+    return E->getMember().getAsString() == MemberName;
+
+  return false;
+}
+
 } // namespace
 
 using utils::isBinaryOrTernary;
@@ -140,9 +159,10 @@ void ContainerSizeEmptyCheck::registerMatchers(MatchFinder *Finder) {
   const auto ValidContainerNonTemplateType =
       qualType(hasUnqualifiedDesugaredType(
           recordType(hasDeclaration(ValidContainerRecord))));
-  const auto ValidContainerTemplateType =
-      qualType(hasUnqualifiedDesugaredType(templateSpecializationType(
-          hasDeclaration(classTemplateDecl(has(ValidContainerRecord))))));
+  const auto ValidContainerTemplateType = qualType(hasUnqualifiedDesugaredType(
+      anyOf(templateSpecializationType(
+                hasDeclaration(classTemplateDecl(has(ValidContainerRecord)))),
+            injectedClassNameType(hasDeclaration(ValidContainerRecord)))));
 
   const auto ValidContainer = qualType(
       anyOf(ValidContainerNonTemplateType, ValidContainerTemplateType));
@@ -155,6 +175,9 @@ void ContainerSizeEmptyCheck::registerMatchers(MatchFinder *Finder) {
                           .bind("SizeBinaryOp")),
             usedInBooleanContext());
 
+  const auto NotInEmptyMethodOfContainer = unless(
+      forCallable(cxxMethodDecl(hasCanonicalDecl(equalsBoundNode("empty")))));
+
   Finder->addMatcher(
       cxxMemberCallExpr(
           argumentCountIs(0),
@@ -164,25 +187,23 @@ void ContainerSizeEmptyCheck::registerMatchers(MatchFinder *Finder) {
                  .bind("MemberCallObject")),
           callee(
               cxxMethodDecl(hasAnyName("size", "length")).bind("SizeMethod")),
-          WrongUse,
-          unless(hasAncestor(
-              cxxMethodDecl(ofClass(equalsBoundNode("container"))))))
+          WrongUse, NotInEmptyMethodOfContainer)
           .bind("SizeCallExpr"),
       this);
 
   Finder->addMatcher(
-      callExpr(argumentCountIs(0),
-               has(cxxDependentScopeMemberExpr(
-                       hasObjectExpression(
-                           expr(anyOf(hasType(ValidContainer),
-                                      hasType(pointsTo(ValidContainer)),
-                                      hasType(references(ValidContainer))))
-                               .bind("MemberCallObject")),
-                       anyOf(hasMemberName("size"), hasMemberName("length")))
-                       .bind("DependentExpr")),
-               WrongUse,
-               unless(hasAncestor(
-                   cxxMethodDecl(ofClass(equalsBoundNode("container"))))))
+      callExpr(
+          argumentCountIs(0),
+          has(mapAnyOf(memberExpr, cxxDependentScopeMemberExpr)
+                  .with(
+                      hasObjectExpression(
+                          expr(anyOf(hasType(ValidContainer),
+                                     hasType(pointsTo(ValidContainer)),
+                                     hasType(references(ValidContainer))))
+                              .bind("MemberCallObject")),
+                      anyOf(matchMemberName("size"), matchMemberName("length")))
+                  .bind("MemberExpr")),
+          WrongUse, NotInEmptyMethodOfContainer)
           .bind("SizeCallExpr"),
       this);
 
@@ -217,8 +238,7 @@ void ContainerSizeEmptyCheck::registerMatchers(MatchFinder *Finder) {
           hasAnyOperatorName("==", "!="), hasOperands(WrongComparend, STLArg),
           unless(allOf(hasLHS(hasType(ExcludedComparisonTypesMatcher)),
                        hasRHS(hasType(SameExcludedComparisonTypesMatcher)))),
-          unless(hasAncestor(
-              cxxMethodDecl(ofClass(equalsBoundNode("container"))))))
+          NotInEmptyMethodOfContainer)
           .bind("BinCmp"),
       this);
 }
@@ -382,9 +402,12 @@ void ContainerSizeEmptyCheck::check(const MatchFinder::MatchResult &Result) {
       Diag << SizeMethod;
     else if (const auto *DependentExpr =
                  Result.Nodes.getNodeAs<CXXDependentScopeMemberExpr>(
-                     "DependentExpr"))
+                     "MemberExpr"))
       Diag << DependentExpr->getMember();
-    else
+    else if (const auto *ME =
+                 Result.Nodes.getNodeAs<MemberExpr>("MemberExpr")) {
+      Diag << ME->getMemberNameInfo().getName();
+    } else
       Diag << "unknown method";
     Diag << Hint;
   } else {