Merge branch 'main' into zgu/profiled_thread

Author: zhengyu123

.github/workflows/ci.yml

@@ -49,7 +49,7 @@ jobs:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
 
       - name: Setup Java
-        uses: actions/setup-java@1bcf9fb12cf4aa7d266a90ae39939e61372fe520 # v5.4.0
+        uses: actions/setup-java@0f481fcb613427c0f801b606911222b5b6f3083a # v5.5.0
         with:
           distribution: 'zulu'
           java-version: '21'
@@ -90,7 +90,7 @@ jobs:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
 
       - name: Setup Java
-        uses: actions/setup-java@1bcf9fb12cf4aa7d266a90ae39939e61372fe520 # v5.4.0
+        uses: actions/setup-java@0f481fcb613427c0f801b606911222b5b6f3083a # v5.5.0
         with:
           distribution: 'zulu'
           java-version: '21'

.github/workflows/codecheck.yml

@@ -76,7 +76,7 @@ jobs:
 
       # Initializes the CodeQL tools for scanning.
       - name: Initialize CodeQL
-        uses: github/codeql-action/init@54f647b7e1bb85c95cddabcd46b0c578ec92bc1a # v4.36.3
+        uses: github/codeql-action/init@99df26d4f13ea111d4ec1a7dddef6063f76b97e9 # v4.37.0
         with:
           languages: ${{ matrix.language }}
           # If you wish to specify custom queries, you can do so here or in a config file.
@@ -85,4 +85,4 @@ jobs:
           # queries: ./path/to/local/query, your-org/your-repo/queries@main
       - run: ./gradlew -x test assembleReleaseJar
       - name: Perform CodeQL Analysis
-        uses: github/codeql-action/analyze@54f647b7e1bb85c95cddabcd46b0c578ec92bc1a # v4.36.3
+        uses: github/codeql-action/analyze@99df26d4f13ea111d4ec1a7dddef6063f76b97e9 # v4.37.0

.github/workflows/release-validated.yml

@@ -43,7 +43,7 @@ jobs:
           fetch-depth: 0
 
       - name: Setup Java
-        uses: actions/setup-java@1bcf9fb12cf4aa7d266a90ae39939e61372fe520 # v5.4.0
+        uses: actions/setup-java@0f481fcb613427c0f801b606911222b5b6f3083a # v5.5.0
         with:
           distribution: 'zulu'
           java-version: '21'
@@ -204,7 +204,7 @@ jobs:
           git checkout $GITHUB_REF_NAME
 
       - name: Setup Java
-        uses: actions/setup-java@1bcf9fb12cf4aa7d266a90ae39939e61372fe520 # v5.4.0
+        uses: actions/setup-java@0f481fcb613427c0f801b606911222b5b6f3083a # v5.5.0
         with:
           distribution: 'zulu'
           java-version: '21'

ddprof-test/src/test/java/com/datadoghq/profiler/cpu/CTimerSamplerTest.java

@@ -72,6 +72,10 @@ protected void after() throws Exception {
 
     @Override
     protected String getProfilerCommand() {
-        return "cpu=100us,event=ctimer";
+        // cpu=100us signal-based sampling is much more expensive under ASAN (the signal
+        // handler itself is ASAN-instrumented), which can inflate the sample count and,
+        // via the per-sample stack-trace materialization above, the test heap. Sample
+        // coarser under ASAN.
+        return isAsan() ? "cpu=1ms,event=ctimer" : "cpu=100us,event=ctimer";
     }
 }

ddprof-test/src/test/java/com/datadoghq/profiler/cpu/LightweightContextCpuTest.java

@@ -79,6 +79,9 @@ protected void after() throws Exception {
 
     @Override
     protected String getProfilerCommand() {
-        return "cpu=100us,lightweight=yes";
+        // cpu=100us signal-based sampling is much more expensive under ASAN (the signal
+        // handler itself is ASAN-instrumented), which can inflate the sample count and,
+        // via the per-sample accessor calls above, the test heap. Sample coarser under ASAN.
+        return isAsan() ? "cpu=1ms,lightweight=yes" : "cpu=100us,lightweight=yes";
     }
 }

ddprof-test/src/test/java/com/datadoghq/profiler/metadata/MetadataNormalisationTest.java

@@ -27,14 +27,18 @@ public void test() throws Exception {
         Method arrayListAdd = ArrayList.class.getDeclaredMethod("add", Object.class);
         Constructor<?> linkedListConstructor = LinkedList.class.getConstructor();
         Method linkedListAdd = LinkedList.class.getDeclaredMethod("add", Object.class);
-        // need to invoke enough times to result in generation of accessors and to record some cpu time
+        // need to invoke enough times to result in generation of accessors and to record some cpu time.
+        // Under ASAN, cpu=100us sampling combined with this fixed iteration count produces a much
+        // larger sample set (stack-trace signal handling is slower under ASAN), and every sample's
+        // stack trace is materialized below, so reduce the iteration count to bound the sample count.
+        int iterations = isAsan() ? 10_000 : 100_000;
         int count = 0;
-        for (int i = 0; i < 100_000; i++) {
+        for (int i = 0; i < iterations; i++) {
             Object list = arrayListConstructor.newInstance();
             arrayListAdd.invoke(list, "element");
             count += ((List<?>) list).size();
         }
-        for (int i = 0; i < 100_000; i++) {
+        for (int i = 0; i < iterations; i++) {
             Object list = linkedListConstructor.newInstance();
             linkedListAdd.invoke(list, "element");
             count += ((List<?>) list).size();
@@ -63,6 +67,6 @@ public void test() throws Exception {
 
     @Override
     protected String getProfilerCommand() {
-        return "cpu=100us";
+        return isAsan() ? "cpu=1ms" : "cpu=100us";
     }
 }

ddprof-test/src/test/java/com/datadoghq/profiler/wallclock/MegamorphicCallTest.java

@@ -20,7 +20,12 @@
 public class MegamorphicCallTest extends AbstractProfilerTest {
     @Override
     protected String getProfilerCommand() {
-        return "wall=100us";
+        // wall=100us over the fixed workload below is ~10k samples/s. Under ASAN each
+        // invocation is much slower, so the same fixed-rate sampling over a much longer
+        // wall-clock duration produces hundreds of thousands of samples, and stack-trace
+        // stringification of all of them (below) OOMs the test-runner heap. Sample 10x
+        // coarser under ASAN, combined with a smaller workload, to bound the sample count.
+        return isAsan() ? "wall=1ms" : "wall=100us";
     }
 
     private static int calculation() {
@@ -60,9 +65,9 @@ public int calculate() {
         }
     }
 
-    private int profiledWork(Calculator... calculators) {
+    private int profiledWork(int iterations, Calculator... calculators) {
         int result = 0;
-        for (int i = 0; i < 1_000_000; i++) {
+        for (int i = 0; i < iterations; i++) {
             for (Calculator calculator : calculators) {
                 result += calculator.calculate();
             }
@@ -74,7 +79,11 @@ private int profiledWork(Calculator... calculators) {
     public void testITableStubs() {
         Assumptions.assumeFalse(Platform.isZing() || Platform.isJ9());
         registerCurrentThreadForWallClockProfiling();
-        int result = profiledWork(new Calculator1(), new Calculator2(), new Calculator3());
+        // Reduce workload under ASAN: combined with the coarser wall rate above, this
+        // bounds the number of samples (and thus the stack-trace strings materialized
+        // below) regardless of how much ASAN slows down each invocation.
+        int iterations = isAsan() ? 100_000 : 1_000_000;
+        int result = profiledWork(iterations, new Calculator1(), new Calculator2(), new Calculator3());
         System.err.println(result);
         stopProfiler();
         IItemCollection events = verifyEvents("datadog.MethodSample");

gradle/libs.versions.toml

@@ -14,7 +14,7 @@ slf4j = "2.0.18"
 # Profiler runtime
 jmc = "8.3.1"
 jol = "0.17"
-lz4 = "1.11.0"
+lz4 = "1.11.1"
 snappy = "1.1.10.8"
 zstd = "1.5.7-11"