Add workflow to post binary size comparison to pull request comments (#1722)

.github/workflows/compare-binary-sizes.yaml

@@ -0,0 +1,28 @@
+name: compare-binary-sizes
+on:
+  workflow_run:
+    workflows: [CI]
+    types: [completed]
+permissions:
+  contents: read
+  pull-requests: write
+  actions: read
+env:
+  GH_TOKEN: "${{ github.token }}"
+  REPO: "${{ github.repository }}"
+  PR_SHA: "${{ github.event.workflow_run.head_sha }}"
+jobs:
+  compare:
+    if: ${{ github.event.workflow_run.event == 'pull_request' }}
+    runs-on: ubuntu-latest
+    steps:
+      - run: |
+          gh api "repos/${REPO}/contents/scripts/compare-binary-sizes.sh" -H 'Accept: application/vnd.github.raw' > compare.sh
+          read pr base_sha < <(gh api "repos/${REPO}/commits/${PR_SHA}/pulls" --jq '.[0] | "\(.number) \(.base.sha)"')
+          function find-archive() { find "${1}" -type f \( -name "async-profiler-*${2}.tar.gz" -o -name "async-profiler-*${2}.zip" \) ! -name '*-debug*' | head -n1; }
+          for p in linux-x64 linux-arm64 macos; do
+            gh run download -R "${REPO}" -n "async-profiler-${p}-${base_sha:0:7}" -D "base/${p}" 2>/dev/null || continue
+            gh run download -R "${REPO}" -n "async-profiler-${p}-${PR_SHA:0:7}"   -D "pr/${p}"   2>/dev/null || continue
+            bash compare.sh "$(find-archive "base/${p}" "${p}")" "$(find-archive "pr/${p}" "${p}")" >> report.md
+          done
+          gh pr comment "${pr}" -R "${REPO}" --body-file report.md --edit-last --create-if-none

CHANGELOG.md

@@ -1,6 +1,6 @@
 # Changelog
 
-## [4.4]
+## [4.4] - 2026-04-20
 
 ### Features
 
@@ -27,15 +27,18 @@
 
 ### Bug fixes
 
+ - #1727: Allocation profile has wrong units in OTLP format
  - #1716: Wall-clock Heatmap does not count samples correctly
  - #1715: Fix Zing crash when profiling cpu+wall together
  - #1708: Another fix for correct vDSO unwinding on ARM64
  - #1707: Workaround for JFR shutdown race
  - #1699: Allow negative keys in JFR constant pool
  - #1697: Ensure remaining buffer is sufficient for event data in JfrReader
+ - #1657: Re-enable workaround for a long attach on JDK 8
  - #1654: Prefer perf-events engine when record-cpu or target-cpu are selected
  - #1585: Scale perf counters in case of multiplexing
  - #1528: Add a hard-coded limit on the maximum number of jmethodIDs
+ - #1203: Fix "Instance field not found" when using `-Xcheck:jni` on JDK 8
  - Do not walk past virtual thread continuation barriers
 
 ## [4.3] - 2026-01-20

README.md

@@ -23,12 +23,12 @@ to learn about more features.
 
 # Download
 
-### Stable release: [4.3](https://github.com/async-profiler/async-profiler/releases/tag/v4.3)
+### Stable release: [4.4](https://github.com/async-profiler/async-profiler/releases/tag/v4.4)
 
-- Linux x64: [async-profiler-4.3-linux-x64.tar.gz](https://github.com/async-profiler/async-profiler/releases/download/v4.3/async-profiler-4.3-linux-x64.tar.gz)
-- Linux arm64: [async-profiler-4.3-linux-arm64.tar.gz](https://github.com/async-profiler/async-profiler/releases/download/v4.3/async-profiler-4.3-linux-arm64.tar.gz)
-- macOS arm64/x64: [async-profiler-4.3-macos.zip](https://github.com/async-profiler/async-profiler/releases/download/v4.3/async-profiler-4.3-macos.zip)
-- Profile converters: [jfr-converter.jar](https://github.com/async-profiler/async-profiler/releases/download/v4.3/jfr-converter.jar)
+- Linux x64: [async-profiler-4.4-linux-x64.tar.gz](https://github.com/async-profiler/async-profiler/releases/download/v4.4/async-profiler-4.4-linux-x64.tar.gz)
+- Linux arm64: [async-profiler-4.4-linux-arm64.tar.gz](https://github.com/async-profiler/async-profiler/releases/download/v4.4/async-profiler-4.4-linux-arm64.tar.gz)
+- macOS arm64/x64: [async-profiler-4.4-macos.zip](https://github.com/async-profiler/async-profiler/releases/download/v4.4/async-profiler-4.4-macos.zip)
+- Profile converters: [jfr-converter.jar](https://github.com/async-profiler/async-profiler/releases/download/v4.4/jfr-converter.jar)
 
 ### Nightly builds
 

scripts/compare-binary-sizes.sh

@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+# Copyright The async-profiler authors
+# SPDX-License-Identifier: Apache-2.0
+
+set -euo pipefail
+
+function list-archive() {
+    case "${1}" in
+        *.tar.gz) tar -tzvf "${1}" --wildcards '*/bin/*' '*/lib/*' | awk '$1 ~ /^-/ { print $NF, $3 }' ;;
+        *.zip)    zipinfo -l "${1}"           '*/bin/*' '*/lib/*'  | awk '$1 ~ /^-/ { print $NF, $4 }' ;;
+    esac | cut -d/ -f2- | sort
+}
+
+function main() {
+    if [[ ${#} -ne 2 ]]; then
+        printf 'Usage: compare-binary-sizes.sh <base-archive> <treatment-archive>\n' >&2
+        exit 1
+    fi
+
+    [[ ${1} =~ (linux-x64|linux-arm64|macos) ]]
+    printf '### %s\n\n| File | Base (KiB) | Treatment (KiB) | Delta (KiB) | %% Change |\n| --- | ---: | ---: | ---: | ---: |\n' "${BASH_REMATCH[1]}"
+    join -a1 -a2 -e0 -o '0,1.2,2.2' \
+         <(list-archive "${1}") \
+         <(list-archive "${2}") \
+        | awk '{
+            b = $2; t = $3; d = t - b; pct = b > 0 ? d * 100 / b : 0
+            printf "| `%s` | %.2f | %.2f | %+.2f | %+.2f%% |\n", $1, b/1024, t/1024, d/1024, pct
+        }'
+}
+
+[[ "${BASH_SOURCE[0]}" == "${0}" ]] && main "$@"

src/api/one/profiler/AsyncProfiler.java

@@ -201,7 +201,7 @@ public class AsyncProfiler implements AsyncProfilerMXBean {
     @Override
     public String dumpCollapsed(Counter counter) {
         try {
-            return execute0("collapsed," + counter.name().toLowerCase());
+            return execute0("collapsed," + (counter == Counter.SAMPLES ? "samples" : "total"));
         } catch (IOException e) {
             throw new IllegalStateException(e);
         }
@@ -242,12 +242,13 @@ public class AsyncProfiler implements AsyncProfilerMXBean {
      * <p>
      * This API is UNSTABLE and might change or be removed in the next version of async-profiler.
      *
+     * @param counter Which counter to use for aggregation
      * @return OTLP representation of the profile
      */
     @Override
-    public byte[] dumpOtlp() {
+    public byte[] dumpOtlp(Counter counter) {
         try {
-            return execute1("otlp");
+            return execute1("otlp," + (counter == Counter.SAMPLES ? "samples" : "total"));
         } catch (IOException e) {
             throw new IllegalStateException(e);
         }

src/api/one/profiler/AsyncProfilerMXBean.java

@@ -31,5 +31,5 @@ public interface AsyncProfilerMXBean {
     String dumpCollapsed(Counter counter);
     String dumpTraces(int maxTraces);
     String dumpFlat(int maxMethods);
-    byte[] dumpOtlp();
+    byte[] dumpOtlp(Counter counter);
 }

src/vmStructs.cpp

@@ -442,7 +442,9 @@ void VMStructs::resolveOffsets() {
         return;
     }
 
-    if (_klass_offset_addr != NULL) {
+    JVMFlag* cjc = JVMFlag::find("CheckJNICalls");
+    if (cjc != NULL && !cjc->get() && _klass_offset_addr != NULL) {
+        // Create a synthetic fieldID to access VMKlass from jclass instance
         _klass = (jfieldID)(uintptr_t)(*_klass_offset_addr << 2 | 2);
     }
 
@@ -461,8 +463,7 @@ void VMStructs::resolveOffsets() {
             && (_compact_object_headers ? (_markword_klass_shift >= 0 && _markword_monitor_value == MONITOR_BIT)
                                         : _oop_klass_offset >= 0)
             && (_symbol_length_offset >= 0 || _symbol_length_and_refcount_offset >= 0)
-            && _symbol_body_offset >= 0
-            && _klass != NULL;
+            && _symbol_body_offset >= 0;
 
     _has_method_structs = _jmethod_ids_offset >= 0
             && _nmethod_method_offset >= 0

test/test/api/DumpOtlp.java

@@ -5,7 +5,6 @@
 
 package test.api;
 
-import java.nio.ByteBuffer;
 import one.profiler.AsyncProfiler;
 import one.profiler.Counter;
 import one.profiler.Events;
@@ -22,7 +21,7 @@ public class DumpOtlp extends BusyLoops {
         }
 
         // TODO: Should we test this further?
-        byte[] profile = AsyncProfiler.getInstance().dumpOtlp();
+        byte[] profile = AsyncProfiler.getInstance().dumpOtlp(Counter.TOTAL);
         System.out.println(profile.length);
     }
 }

test/test/otlp/OtlpProfileTimeTest.java

@@ -5,6 +5,7 @@
 package test.otlp;
 
 import one.profiler.AsyncProfiler;
+import one.profiler.Counter;
 import one.profiler.Events;
 import io.opentelemetry.proto.profiles.v1development.*;
 
@@ -39,7 +40,7 @@ public class OtlpProfileTimeTest {
     }
 
     public static Profile dumpAndGetProfile(AsyncProfiler profiler) throws Exception {
-        byte[] dump = profiler.dumpOtlp();
+        byte[] dump = profiler.dumpOtlp(Counter.SAMPLES);
         ProfilesData data = ProfilesData.parseFrom(dump);
         return data.getResourceProfiles(0).getScopeProfiles(0).getProfiles(0);
     }

test/test/vmstructs/VmstructsTests.java

@@ -5,9 +5,11 @@
 
 package test.vmstructs;
 
+import one.profiler.test.Assert;
 import one.profiler.test.Output;
 import one.profiler.test.Test;
 import one.profiler.test.TestProcess;
+import test.alloc.Hello;
 import test.smoke.Alloc;
 
 public class VmstructsTests {
@@ -25,4 +27,10 @@ public class VmstructsTests {
         Output out = p.profile("stop -o collapsed");
         assert out.contains("Alloc.allocate");
     }
+
+    @Test(mainClass = Hello.class, jvmArgs = "-Xcheck:jni", agentArgs = "start,event=alloc", output = true)
+    public void checkJni(TestProcess p) throws Exception {
+        p.waitForExit();
+        Assert.isEqual(p.exitCode(), 0);
+    }
 }