ci(framework): bound linux workers and retain test diagnostics

.github/actions/upload-test-diagnostics/action.yml

@@ -0,0 +1,36 @@
+name: Upload test diagnostics
+description: Preserve test reports, application logs, and JVM crash reports after a failed build.
+
+runs:
+  using: composite
+  steps:
+    - name: Report available memory and process limits
+      if: always()
+      shell: bash
+      run: |
+        ulimit -a
+        if command -v free >/dev/null 2>&1; then
+          free -m
+        elif command -v vm_stat >/dev/null 2>&1; then
+          vm_stat
+        fi
+        for counter in /sys/fs/cgroup/memory.events /sys/fs/cgroup/memory.current /sys/fs/cgroup/memory.max; do
+          if [ -r "$counter" ]; then
+            echo "$counter"
+            cat "$counter"
+          fi
+        done
+
+    - name: Upload diagnostic files
+      if: always()
+      uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
+      with:
+        name: test-diagnostics-${{ github.job }}-${{ github.run_attempt }}
+        path: |
+          **/build/test-results/**
+          **/build/reports/tests/**
+          **/logs/tron-test*.log*
+          **/hs_err_pid*.log
+        if-no-files-found: warn
+        retention-days: 7
+        compression-level: 1

.github/workflows/pr-build.yml

@@ -63,6 +63,10 @@ jobs:
           java -jar "$JAR" db archive -h
           java -jar "$JAR" keystore --help
 
+      - name: Upload test diagnostics
+        if: failure()
+        uses: ./.github/actions/upload-test-diagnostics
+
   build-ubuntu:
     name: Build ubuntu24 (JDK 17 / aarch64)
     if: ${{ github.event_name == 'pull_request' || inputs.job == 'all' || inputs.job == 'ubuntu' }}
@@ -91,7 +95,9 @@ jobs:
           restore-keys: ubuntu24-aarch64-gradle-
 
       - name: Build
-        run: ./gradlew clean build --no-daemon
+        # Bound worker concurrency while retaining parallelism within the 60-minute job limit.
+        # Execute tests on every retry instead of restoring their outputs from the build cache.
+        run: ./gradlew clean build --no-daemon --max-workers=2 --rerun-tasks --info --stacktrace
 
       - name: Toolkit jar smoke test
         run: |
@@ -102,6 +108,10 @@ jobs:
           java -jar "$JAR" db archive -h
           java -jar "$JAR" keystore --help
 
+      - name: Upload test diagnostics
+        if: failure()
+        uses: ./.github/actions/upload-test-diagnostics
+
   docker-build-rockylinux:
     name: Build rockylinux (JDK 8 / x86_64)
     if: ${{ github.event_name == 'pull_request' || inputs.job == 'all' || inputs.job == 'rockylinux' }}
@@ -143,7 +153,9 @@ jobs:
         run: ./gradlew --stop || true
 
       - name: Build
-        run: ./gradlew clean build --no-daemon
+        # Bound worker concurrency while retaining parallelism within the 60-minute job limit.
+        # Execute tests on every retry instead of restoring their outputs from the build cache.
+        run: ./gradlew clean build --no-daemon --max-workers=2 --rerun-tasks --info --stacktrace
 
       - name: Toolkit jar smoke test
         run: |
@@ -155,7 +167,11 @@ jobs:
           java -jar "$JAR" keystore --help
 
       - name: Test with RocksDB engine
-        run: ./gradlew :framework:testWithRocksDb --no-daemon
+        run: ./gradlew :framework:testWithRocksDb --no-daemon --max-workers=2 --rerun-tasks --info --stacktrace
+
+      - name: Upload test diagnostics
+        if: failure()
+        uses: ./.github/actions/upload-test-diagnostics
 
   docker-build-debian11:
     name: Build debian11 (JDK 8 / x86_64)
@@ -222,6 +238,10 @@ jobs:
             **/build/reports/jacoco/test/jacocoTestReport.xml
           if-no-files-found: error
 
+      - name: Upload test diagnostics
+        if: failure()
+        uses: ./.github/actions/upload-test-diagnostics
+
   coverage-base:
     name: Coverage Base (JDK 8 / x86_64)
     if: ${{ github.event_name == 'pull_request' }}

framework/build.gradle

@@ -136,6 +136,10 @@ def configureTestTask = { Task t ->
     }
     t.maxHeapSize = "512m"
     t.maxParallelForks = Math.max(1, Math.min(4, Runtime.runtime.availableProcessors()))
+    if ('true'.equalsIgnoreCase(System.getenv('CI'))) {
+        // Keep fatal test-JVM diagnostics in the task output directory for CI artifacts.
+        t.jvmArgs "-XX:ErrorFile=${t.temporaryDir}/hs_err_pid%p.log"
+    }
     t.systemProperty 'runPrecompileBenchmark',
             System.getProperty('runPrecompileBenchmark', 'false')
     t.doFirst {

framework/src/test/java/org/tron/core/services/ComputeRewardTest.java

@@ -10,7 +10,9 @@
 import java.util.List;
 import java.util.Map;
 import java.util.concurrent.ExecutionException;
+import java.util.concurrent.ExecutorService;
 import java.util.concurrent.Future;
+import java.util.concurrent.TimeUnit;
 import org.junit.Assert;
 import org.junit.Rule;
 import org.junit.Test;
@@ -248,8 +250,15 @@ private void setUp() {
   }
 
   @Test
-  public void query() {
+  public void query() throws InterruptedException {
     Assert.assertEquals(3189, mortgageService.queryReward(OWNER_ADDRESS));
+    // queryReward() can return before the background Merkle-root calculation finishes.
+    // Wait for it before modifying or closing the store used by that calculation.
+    ExecutorService rewardExecutor = ReflectUtils.getFieldValue(rewardViCalService, "es");
+    rewardExecutor.shutdown();
+    Assert.assertTrue("Reward calculation did not terminate before modifying its store",
+        rewardExecutor.awaitTermination(10, TimeUnit.SECONDS));
+
     // mock root is error
     rewardViStore.put("test".getBytes(), "test".getBytes());
     ReflectUtils.invokeMethod(rewardViCalService,"maybeRun");