以前，我曾說過， 您總是想保留一些調試器斷點作為例外 。 此幫助可防止代碼在不引起注意的情況下腐爛掉-有時掩蓋了另一個問題。

如果您認真對待這一點，則最好將此概念擴展到自動化測試中。 但是想出一個全面的解決方案并不簡單。 您可以僅從try / catch開始，但不會捕獲其他線程上的異常。 您還可以使用AOP進行操作； 但是，根據框架的不同，您不能保證完全捕獲所有內容，這確實意味著您正在使用稍微不同的代碼進行測試，這將使您有些擔憂。

幾天前，我遇到了有關如何編寫自己的調試器的博客文章，我想知道java進程是否有可能自行調試。 事實證明，可以的，這是我作為這個小小的思想實驗的一部分提出的代碼。

該類的第一部分僅包含一些相當hacky的代碼，用于根據啟動參數來猜測連接回同一VM所需的端口。 可能可以使用Attach機制啟動調試器。 但是我沒有看到一種明顯的方法來使其工作。 然后只有幾個工廠方法帶有要查找的異常列表。

package com.kingsfleet.debug;import com.sun.jdi.Bootstrap; import com.sun.jdi.ReferenceType; import com.sun.jdi.VirtualMachine; import com.sun.jdi.connect.AttachingConnector; import com.sun.jdi.connect.Connector; import com.sun.jdi.connect.IllegalConnectorArgumentsException; import com.sun.jdi.event.ClassPrepareEvent; import com.sun.jdi.event.Event; import com.sun.jdi.event.EventQueue; import com.sun.jdi.event.EventSet; import com.sun.jdi.event.ExceptionEvent; import com.sun.jdi.event.VMDeathEvent; import com.sun.jdi.event.VMDisconnectEvent; import com.sun.jdi.request.ClassPrepareRequest; import com.sun.jdi.request.EventRequest; import com.sun.jdi.request.ExceptionRequest;import java.io.IOException;import java.lang.management.ManagementFactory; import java.lang.management.RuntimeMXBean;import java.util.Collections; import java.util.HashSet; import java.util.List; import java.util.Map; import java.util.Set; import java.util.concurrent.CountDownLatch; import java.util.concurrent.TimeUnit;public class ExceptionDebugger implements AutoCloseable {public static int getDebuggerPort() {// Try to work out what port we need to connect toRuntimeMXBean runtime = ManagementFactory.getRuntimeMXBean();List<String> inputArguments = runtime.getInputArguments();int port = -1;boolean isjdwp = false;for (String next : inputArguments) {if (next.startsWith("-agentlib:jdwp=")) {isjdwp = true;String parameterString = next.substring("-agentlib:jdwp=".length());String[] parameters = parameterString.split(",");for (String parameter : parameters) {if (parameter.startsWith("address")) {int portDelimeter = parameter.lastIndexOf(":");if (portDelimeter != -1) {port = Integer.parseInt(parameter.substring(portDelimeter + 1));} else {port = Integer.parseInt(parameter.split("=")[1]);}}}}}return port;}public static ExceptionDebugger connect(final String... exceptions) throws InterruptedException {return connect(getDebuggerPort(),exceptions);}public static ExceptionDebugger connect(final int port, final String... exceptions) throws InterruptedException {ExceptionDebugger ed = new ExceptionDebugger(port, exceptions);return ed;}

構造函數創建一個簡單的守護程序線程，以啟動與虛擬機的連接。 這是一個單獨的線程，這一點非常重要，否則顯然，當我們遇到斷點時，VM會停止運行。 確保該線程中的代碼不會引發異常是一個好主意-目前，我只是希望做到最好。

最后，代碼僅維護了一個禁止的異常列表，如果您有更多的時間，應該可以在發生異常的地方存儲堆棧跟蹤。

// // Instance variablesprivate final CountDownLatch startupLatch = new CountDownLatch(1);private final CountDownLatch shutdownLatch = new CountDownLatch(1);private final Set<String> set = Collections.synchronizedSet(new HashSet<String>());private final int port;private final String exceptions[];private Thread debugger;private volatile boolean shutdown = false;//// Object construction and methods//private ExceptionDebugger(final int port, final String... exceptions) throws InterruptedException {this.port = port;this.exceptions = exceptions;debugger = new Thread(new Runnable() {@Overridepublic void run() {try {connect();} catch (Exception ex) {ex.printStackTrace();}}}, "Self debugging");debugger.setDaemon(true); // Don't hold the VM opendebugger.start();// Make sure the debugger has connectedif (!startupLatch.await(1, TimeUnit.MINUTES)) {throw new IllegalStateException("Didn't connect before timeout");}}@Overridepublic void close() throws InterruptedException {shutdown = true;// Somewhere in JDI the interrupt was being eaten, hence the volatile flag debugger.interrupt();shutdownLatch.await();}/*** @return A list of exceptions that were thrown*/public Set<String> getExceptionsViolated() {return new HashSet<String>(set);}/*** Clear the list of exceptions violated*/public void clearExceptionsViolated() {set.clear();}

主要的connect方法是一個相當簡單的代碼塊，可確保連接并配置任何初始斷點。

//// Implementation details//private void connect() throws java.io.IOException {try {// Create a virtual machine connectionVirtualMachine attach = connectToVM();try{// Add prepare and any already loaded exception breakpointscreateInitialBreakpoints(attach);// We can now allow the rest of the work to go on as we have created the breakpoints// we requiredstartupLatch.countDown();// Process the eventsprocessEvents(attach);}finally {// Disconnect the debuggerattach.dispose();// Give the debugger time to really disconnect// before we might reconnect, couldn't find another// way to do thistry {TimeUnit.SECONDS.sleep(1);} catch (InterruptedException e) {Thread.currentThread().interrupt();}}} finally {// Notify watchers that we have shutdownshutdownLatch.countDown();}}

重新連接到自我只是找到合適的連接器的過程，在本例中是Socket，盡管我猜想如果您稍稍修改一下代碼，便可以在某些平臺上使用共享內存傳輸。

private VirtualMachine connectToVM() throws java.io.IOException {List<AttachingConnector> attachingConnectors = Bootstrap.virtualMachineManager().attachingConnectors();AttachingConnector ac = null;found:for (AttachingConnector next : attachingConnectors) {if (next.name().contains("SocketAttach")) {ac = next;break;}}Map<String, Connector.Argument> arguments = ac.defaultArguments();arguments.get("hostname").setValue("localhost");arguments.get("port").setValue(Integer.toString(port));arguments.get("timeout").setValue("4000");try {return ac.attach(arguments);} catch (IllegalConnectorArgumentsException e) {throw new IOException("Problem connecting to debugger",e);}}

連接調試器時，您不知道是否已加載您感興趣的異常，因此您需要為準備類的點和已經加載的點注冊斷點。

請注意，設置的斷點僅用于斷開一個線程的策略–否則，出于顯而易見的原因，如果調試器線程也進入睡眠狀態，則當前VM將停止運行。

private void createInitialBreakpoints(VirtualMachine attach) {// Our first exception is for class loadingfor (String exception : exceptions) {ClassPrepareRequest cpr = attach.eventRequestManager().createClassPrepareRequest();cpr.addClassFilter(exception);cpr.setSuspendPolicy(EventRequest.SUSPEND_EVENT_THREAD);cpr.setEnabled(true);}// Then we can check each in turn to see if it have already been loaded as we might// be late to the game, remember classes can be loaded more than once//for (String exception : exceptions) {List<ReferenceType> types = attach.classesByName(exception);for (ReferenceType type : types) {createExceptionRequest(attach, type);}}}private static void createExceptionRequest(VirtualMachine attach, ReferenceType refType) {ExceptionRequest er = attach.eventRequestManager().createExceptionRequest(refType, true, true);er.setSuspendPolicy(EventRequest.SUSPEND_EVENT_THREAD);er.setEnabled(true);}

事件處理循環輪詢包含一個或多個事件實例的EventSet實例。 盡管并非所有這些事件都屬于斷點請求，所以您必須注意不要總是在事件集上調用簡歷。 這是因為您可能連續有兩個事件集，而代碼甚至在閱讀第二個事件集之前就調用了resume。 隨著代碼的趕上，這會導致錯過斷點。

出于某種原因， JDI似乎正在吃掉中斷的標志，因此該布爾屬性使用以前的close方法停止循環。

private void processEvents(VirtualMachine attach) {// Listen for eventsEventQueue eq = attach.eventQueue();eventLoop: while (!Thread.interrupted() && !shutdown) {// Poll for event sets, with a short timeout so that we can// be interrupted if requiredEventSet eventSet = null;try {eventSet = eq.remove(500);}catch (InterruptedException ex) {Thread.currentThread().interrupt();continue eventLoop; }// Just loop again if we have no eventsif (eventSet == null) {continue eventLoop;}//boolean resume = false;for (Event event : eventSet) {EventRequest request = event.request();if (request != null) {int eventPolicy = request.suspendPolicy();resume |= eventPolicy != EventRequest.SUSPEND_NONE;}if (event instanceof VMDeathEvent || event instanceof VMDisconnectEvent) {// This should never happen as the VM will exit before this is called} else if (event instanceof ClassPrepareEvent) {// When an instance of the exception class is loaded attach an exception breakpointClassPrepareEvent cpe = (ClassPrepareEvent) event;ReferenceType refType = cpe.referenceType();createExceptionRequest(attach, refType);} else if (event instanceof ExceptionEvent) {String name = ((ExceptionRequest)event.request()).exception().name();set.add(name);}}// Dangerous to call resume always because not all event suspend the VM// and events happen asynchornously.if (resume)eventSet.resume();}}}

因此，剩下的只是一個簡單的測試示例，因為這是JDK 7，而ExceptionDebugger是AutoCloseable，我們可以使用try-with-resources構造進行此操作，如下所示。 顯然，如果要進行自動化測試，請使用您選擇的測試框架固定裝置。

public class Target {public static void main(String[] args) throws InterruptedException {try (ExceptionDebugger ex = ExceptionDebugger.connect(NoClassDefFoundError.class.getName())) {doSomeWorkThatQuietlyThrowsAnException();System.out.println(ex.getExceptionsViolated());}System.exit(0);}private static void doSomeWorkThatQuietlyThrowsAnException() {// Check to see that break point gets firedtry {Thread t = new Thread(new Runnable() {public void run() {try{throw new NoClassDefFoundError();}catch (Throwable ex) {}}});t.start();t.join();} catch (Throwable th) {// Eat this and don't tell anybody}} }

因此，如果使用以下VM參數運行此類，請注意suspend = n，否則代碼將不會開始運行，您會發現它可以重新連接到自身并開始運行。

-agentlib:jdwp=transport=dt_socket,address=localhost:5656,server=y,suspend=n

這將為您提供以下輸出，請注意來自VM的額外調試行：

Listening for transport dt_socket at address: 5656java.lang.NoClassDefFoundError Listening for transport dt_socket at address: 5656

每個人都想讀一下這是否對人們有用并有助于消除任何明顯的錯誤。

參考：在Gerard Davison的博客博客中，從JCG合作伙伴 Gerard Davison 編寫了一個自動調試器，以在測試執行期間捕獲異常 。

翻譯自: https://www.javacodegeeks.com/2013/10/write-an-auto-debugger-to-catch-exceptions-during-test-execution.html

總結

以上是生活随笔為你收集整理的编写自动调试器以在测试执行期间捕获异常的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。