improved thread death error diagnostics.

git-svn-id: https://hudson.dev.java.net/svn/hudson/trunk/hudson/main@4893 71c3de6d-444a-0410-be80-ed276b4c234a

core/src/main/java/hudson/Functions.java

@@ -27,6 +27,8 @@ import javax.servlet.http.HttpServletRequest;
 import javax.servlet.http.HttpServletResponse;
 import java.io.File;
 import java.io.IOException;
+import java.io.StringWriter;
+import java.io.PrintWriter;
 import java.lang.management.LockInfo;
 import java.lang.management.ManagementFactory;
 import java.lang.management.MonitorInfo;
@@ -622,4 +624,10 @@ public class Functions {
         if(b==null) return true;
         return b;
     }
+
+    public static String printThrowable(Throwable t) {
+        StringWriter sw = new StringWriter();
+        t.printStackTrace(new PrintWriter(sw));
+        return sw.toString();
+    }
 }

core/src/main/java/hudson/model/Executor.java

@@ -13,7 +13,7 @@ import java.io.IOException;
  *
  * @author Kohsuke Kawaguchi
  */
-public class Executor extends Thread {
+public class Executor extends Thread implements ModelObject {
     private final Computer owner;
     private final Queue queue;
 
@@ -25,6 +25,8 @@ public class Executor extends Thread {
     private int number;
     private Queue.Executable executable;
 
+    private Throwable causeOfDeath;
+
     public Executor(Computer owner) {
         super("Executor #"+owner.getExecutors().size()+" for "+owner.getDisplayName());
         this.owner = owner;
@@ -34,36 +36,44 @@ public class Executor extends Thread {
     }
 
     public void run() {
-        while(true) {
-            if(Hudson.getInstance().isTerminating())
-                return;
-
-            synchronized(owner) {
-                if(owner.getNumExecutors()<owner.getExecutors().size()) {
-                    // we've got too many executors.
-                    owner.removeExecutor(this);
+        try {
+            while(true) {
+                if(Hudson.getInstance().isTerminating())
                     return;
+
+                synchronized(owner) {
+                    if(owner.getNumExecutors()<owner.getExecutors().size()) {
+                        // we've got too many executors.
+                        owner.removeExecutor(this);
+                        return;
+                    }
                 }
-            }
 
-            Queue.Task task;
-            try {
-                task = queue.pop();
-            } catch (InterruptedException e) {
-                continue;
-            }
+                Queue.Task task;
+                try {
+                    task = queue.pop();
+                } catch (InterruptedException e) {
+                    continue;
+                }
 
-            try {
-                startTime = System.currentTimeMillis();
-                executable = task.createExecutable();
-                queue.execute(executable,task);
-            } catch (Throwable e) {
-                // for some reason the executor died. this is really
-                // a bug in the code, but we don't want the executor to die,
-                // so just leave some info and go on to build other things
-                e.printStackTrace();
+                try {
+                    startTime = System.currentTimeMillis();
+                    executable = task.createExecutable();
+                    queue.execute(executable,task);
+                } catch (Throwable e) {
+                    // for some reason the executor died. this is really
+                    // a bug in the code, but we don't want the executor to die,
+                    // so just leave some info and go on to build other things
+                    e.printStackTrace();
+                }
+                executable = null;
             }
-            executable = null;
+        } catch(RuntimeException e) {
+            causeOfDeath = e;
+            throw e;
+        } catch (Error e) {
+            causeOfDeath = e;
+            throw e;
         }
     }
 
@@ -77,6 +87,13 @@ public class Executor extends Thread {
         return executable;
     }
 
+    /**
+     * Same as {@link #getName()}.
+     */
+    public String getDisplayName() {
+        return "Executor #"+getNumber();
+    }
+
     /**
      * Gets the executor number that uniquely identifies it among
      * other {@link Executor}s for the same computer.
@@ -95,6 +112,16 @@ public class Executor extends Thread {
         return executable==null;
     }
 
+    /**
+     * If this thread dies unexpectedly, obtain the cause of the failure.
+     *
+     * @return null if the death is expected death or the thread is {@link #isAlive() still alive}.
+     * @since 1.142
+     */
+    public Throwable getCauseOfDeath() {
+        return causeOfDeath;
+    }
+
     /**
      * Returns the progress of the current build in the number between 0-100.
      *

core/src/main/resources/hudson/model/Executor/causeOfDeath.jelly

+<j:jelly xmlns:j="jelly:core" xmlns:st="jelly:stapler" xmlns:d="jelly:define" xmlns:l="/lib/layout" xmlns:t="/lib/hudson" xmlns:f="/lib/form" xmlns:i="jelly:fmt">
+  <l:layout title="${it.name}">
+    <l:header />
+    <l:side-panel>
+      <l:tasks>
+        <l:task icon="images/24x24/up.gif" href="../.." title="Back" />
+      </l:tasks>
+    </l:side-panel>
+    <l:main-panel>
+      <j:choose>
+        <j:when test="${it.alive}">
+          <h1>
+            <img src="${imagesURL}/48x48/blue.gif" width="48" height="48" />
+            Thread is still alive
+          </h1>
+        </j:when>
+        <j:otherwise>
+          <h1>
+            <img src="${imagesURL}/48x48/red.gif" width="48" height="48" />            
+            Thread has died
+          </h1>
+          <pre>${h.printThrowable(it.causeOfDeath)}</pre>
+          <pre>
+            <a href="http://hudson.gotdns.com/wiki/display/HUDSON/Dead+Executor">more info</a>
+          </pre>
+        </j:otherwise>
+      </j:choose>
+    </l:main-panel>
+  </l:layout>
+</j:jelly>

core/src/main/resources/lib/hudson/executors.jelly

@@ -40,7 +40,7 @@
 	        <j:choose>
             <j:when test="${!e.alive}">
               <td class="pane" width="70%">
-                <a href="http://hudson.gotdns.com/wiki/display/HUDSON/Dead+Executor">
+                <a href="${rootURL}/computers/${cloop.index}/executors/${eloop.index}/causeOfDeath">
                   <span class="error">Dead (!)</span>
                 </a>
               </td>