Community
Participate
Working Groups
Hi, There must be a bug somewhere in there on Windows 7 at least. This happens only when running SWTBot in headless mode. The effect is that the internal web browser is not available. The code below shows why this is not available: because of WebSite.createTempStorage() -> COM.StgCreateDocfile(...) fails with STG_E_INVALIDNAME. I also ran it "as Administrator" to make sure there are no permission problems. But the behaviour is the same. I have one test containing the lines below. I run this test with SWTBot in headless mode: public void test_InternalBrowser() throws Exception { final StringBuffer sb = new StringBuffer(); PlatformUI.getWorkbench().getDisplay().syncExec(new Runnable() { public void run() { sb.append("canUseInternalWebBrowser: " + WebBrowserUtil.canUseInternalWebBrowser() + "\n"); } }); try { Class.forName("org.eclipse.swt.browser.Browser"); sb.append("class \"org.eclipse.swt.browser.Browser\" found.\n"); } catch (ClassNotFoundException e) { sb.append("class \"org.eclipse.swt.browser.Browser\" not found.\n"); } assertTrue(sb.toString(), false); } The output is: canUseInternalWebBrowser: false class "org.eclipse.swt.browser.Browser" found The problem is that WebBrowserUtil.canUseInternalWebBrowser() tries to construct a Browser object. The call goes down to org.eclipse.swt.browser.WebSite.createTempStorage() method inherited from org.eclipse.swt.ole.win32.OleClientSite: protected IStorage createTempStorage() { int /*long*/[] tempStorage = new int /*long*/[1]; int grfMode = COM.STGM_READWRITE | COM.STGM_SHARE_EXCLUSIVE | COM.STGM_DELETEONRELEASE; int result = COM.StgCreateDocfile(null, grfMode, 0, tempStorage); if (result != COM.S_OK) OLE.error(OLE.ERROR_CANNOT_CREATE_FILE, result); return new IStorage(tempStorage[0]); } After the call to the native method COM.StgCreateDocfile(...) result has the value -2147286788. This translates to 0x800300FC which is STG_E_INVALIDNAME http://msdn.microsoft.com/en-us/library/aa380323(VS.85).aspx 1. Why is the name not valid if NULL is also a valid input value as the page above describes? Was there missing in the initialization of Eclipse/SWT (reponsability of SWTBot I guess)? 2. Eclipse calls this function many times and it always succeeds called in the same way (providing NULL as the first parameter). Is also succeeds when SWTBot is hosted in Eclipse (the framework/SWT/environment/etc. is already set up by the Eclipse host. But in headless mode, when SWTBot starts Eclipse (using the application: org.eclipse.swtbot.eclipse.junit4.headless.swtbottestapplication), something is missing. Any idea what is going wrong here? Thanks! Gabriel
If this can help, Eclipse was started like this: [echo] SWTBot tests fileset: Setup scanner in dir D:\s\plugins with patternSet{ includes: [org.eclipse.equinox.launcher_*.jar] excludes: [] } [java] Executing 'C:\Program Files\Java\jre6\bin\java.exe' with arguments: [java] '-Xms256M' [java] '-Xmx768M' [java] '-XX:MaxPermSize=512M' [java] '-Xdebug' [java] '-Xrunjdwp:transport=dt_socket,server=y,suspend=n,address=1044' [java] '-classpath' [java] 'D:\s\plugins\org.eclipse.equinox.launcher_1.1.0.v20100507.jar' [java] 'org.eclipse.core.launcher.Main' [java] '-application' [java] 'org.eclipse.swtbot.eclipse.junit4.headless.swtbottestapplication' [java] '-data' [java] 'D:\s/test-workspace' [java] 'formatter=org.apache.tools.ant.taskdefs.optional.junit.XMLJUnitResultFormatter,D:\s/result.xml' [java] 'formatter=org.apache.tools.ant.taskdefs.optional.junit.PlainJUnitResultFormatter' [java] '-testPluginName' [java] 'mt.plugin.to.test' [java] '-className' [java] 'my.test.plugin.AllTests' [java] '-os' [java] 'win32' [java] '-ws' [java] 'win32' [java] '-arch' [java] 'x86' [java] '-consoleLog' [java] '-debug' And my.test.plugin.AllTests contains: public static Test suite() { TestSuite suite = new TestSuite(); suite.addTestSuite(InitTestCase.class); return suite; } and InitTestCase contains the test shown in the previous comment: public void test_InternalBrowser() throws Exception { // from previous comment }
More links: if you wonder what "org.eclipse.swtbot.eclipse.junit4.headless.swtbottestapplication" is doing, here is the code: http://dev.eclipse.org/viewcvs/index.cgi/trunk/org.eclipse.swtbot.eclipse.junit4.headless/src/org/eclipse/swtbot/eclipse/junit4/headless/UITestApplication.java?view=markup&root=Technology_SWTBot
This happens also in 3.6.1
Created attachment 179585 [details] Instructions to reproduce I have attached a zip file to reproduce the problem on Windows 7 with Eclipse 3.6.1
Added Ketan from SWTBot to this bug CC.
On XP this seems to work for me as expected. On Windows 7 the launch.bat runs to completion but eclipse is never launched, so perhaps something's a bit wrong with my setup on that machine. Regardless, the success on XP indicates that the problem you see on Windows 7 is likely an swt issue.
Make sure to pick the ant in the org.apache.ant plug-in to make sure we have the same scenarios.
(In reply to comment #6) > On XP this seems to work for me as expected. On Windows 7 the launch.bat runs > to completion but eclipse is never launched, so perhaps something's a bit wrong > with my setup on that machine. Regardless, the success on XP indicates that > the problem you see on Windows 7 is likely an swt issue. Or maybe you have jre6_21 installed? (Known problem on Windows) http://wiki.eclipse.org/FAQ_How_do_I_run_Eclipse%3F (Notice the note in the first section) In this case you have to revert to jre6_20
Created attachment 182180 [details] New Instructions to reproduce Added a slightly updated sample code. Just follow the steps in the README file. This happens only in Windows 7. I removed the debugging JVM and ant flags. I tested it in Windows Vista and Windows XP. Everything works just fine. But in Windows 7 you must get in the console the failed assertion below. I have tested this with Eclipse 3.6 and 3.6.1 on two different Windows 7 machines. Every time the same: default: [java] Exception in thread "WorkbenchTestable" org.eclipse.core.runtime.AssertionFailedException: assertion failed: Cannot use internal Web browser. [java] at org.eclipse.core.runtime.Assert.isTrue(Assert.java:110) [java] at my.test.app.MyApp.runTests(MyApp.java:72) [java] at org.eclipse.ui.internal.testing.WorkbenchTestable$1.run(WorkbenchTestable.java:71) [java] at java.lang.Thread.run(Unknown Source) Anyone, any news on this?
Can someone give this a try on a Windows 7 machine? It should be easy: launch.bat path\to\eclipse wait about 5-10 seconds until eclipse starts. 1. Does the Wellcome page look ok? if white and all icons aligned in a central line, than that's the problem 2. Ctrl-3 -> Web Broser -> enter if you see an error in the browser view, that's the problem again 3. Preferences -> search for Browser if "Use internal web browser" is disabled, that's the problem once again everything happens because of OleClientSite:543: int result = COM.StgCreateDocfile(null, grfMode, 0, tempStorage); result has a big weird value instead of 0 (COM.S_OK) :( we are pretty much blocked on this Thanks for help!
I got your scenario to run on our Windows 7 machine (your plug-in was previously not resolving because the installed jre was 1.5.0). However I do not see the problem. Eclipse now comes up for me, but Welcome page shows its HTML content fine (not the white screen), and the internal web browser view works as well. So there's presumably a difference between our machines. If the Browser works for you when not in headless mode, then as you say there could be some initialization missing. createTempStorage() is one of the first calls made on the site, so it may just happen to be the first place where a bad state shows through. I assume that test_InternalBrowser() is run in a non-UI thread, hence the syncExec() it contains. OleInitialize(0) is called when a Display is created, but I wonder if the Display may have been created on the wrong thread (OleInitialize(0) needs to be called on each COM apartment). I'm going to ask SSQ about this tomorrow, but as a quick experiment in the meantime, can you add "OS.OleInitialize(0)" to your snippet just before the call to canUseInternalWebBrowser() ? I'm not sure if it will help, but it may at least rule out one possibility.
(In reply to comment #11) > I assume that test_InternalBrowser() is run in a non-UI thread, hence the > syncExec() it contains. OleInitialize(0) is called when a Display is created, > but I wonder if the Display may have been created on the wrong thread > (OleInitialize(0) needs to be called on each COM apartment). I'm going to ask > SSQ about this tomorrow, but as a quick experiment in the meantime, can you add > "OS.OleInitialize(0)" to your snippet just before the call to > canUseInternalWebBrowser() ? I'm not sure if it will help, but it may at least > rule out one possibility. A call to OleInitialize(0) does not help. The welcome page still messed-up (see the attached png). Internal web browser not available.
Created attachment 182693 [details] The wrongly displayed Eclipse Wellcome page
Created attachment 182695 [details] The Internal browser is missing
Finally, I found the problem (not the solution): Execute this before calling launch.bat SET TMP=non_existing_directory_or_empty If TMP is not pointing to an existing directory you will get the behaviour described. If this is a bug or not, I don't know. Can you please confirm that this is the case also for you? And is this a bug? Thanks!
Yes, I can confirm that I see this as well (thanks for figuring this out). I guess it makes sense given where the failure is occuring. I've talked to SSQ about this, and it's not an eclipse/swt bug since it's caused by a user setting that's not valid. Closing.
