The next patch I will upload contains the fixes I have indicated as done. I also did not review carefully all the broken tests. I mostly mechanically put the annotation just to get the whole test suite to pass.
Currently, batching will not work with htmlUnit because the batching code does not yet understand the @DoNotRunWith annotation. When returning the batch of tests to execute, the batching code should filter out tests that are annotated @DoNotRunWith. I felt the fix could be postponed till later. Regarding having more specific platforms, I chose the simple HtmlUnit vs non currently, because currently the test-runner runs the same tests for all specified clients. There is no currently no support for having arguments "-htmlunit FF3,IE7" and having a test method marked as @DoNotRunWith(Platform.Htmlunit_IE7). Again, this is a TODO but till then, it will not be useful to have a separate Htmlunit_IE7 and Htmlunit_FF3. http://gwt-code-reviews.appspot.com/54809/diff/38/1001 File eclipse/user/.classpath (right): http://gwt-code-reviews.appspot.com/54809/diff/38/1001#newcode6 Line 6: <classpathentry kind="con" path="org.eclipse.jdt.launching.JRE_CONTAINER"/> Will fix in the next patch. On 2009/07/28 23:36:55, jat wrote: > Apparently the diff didn't match what was in trunk -- see the error: old chunk > mismatch http://gwt-code-reviews.appspot.com/54809/diff/38/1004 File user/src/com/google/gwt/junit/JUnitShell.java (right): http://gwt-code-reviews.appspot.com/54809/diff/38/1004#newcode238 Line 238: return runStyle != null; Changing it to true. On 2009/07/28 23:08:41, scottb wrote: > We're not doing any validation on the input. It's an assertion in > RunStyleHtmlUnit, and the return value from a new op can never be null. http://gwt-code-reviews.appspot.com/54809/diff/38/1004#newcode398 Line 398: private static final int TEST_BEGIN_TIMEOUT_MILLIS = 6000000; This is a mistake. Will change this. On 2009/07/28 23:08:41, scottb wrote: > Seems unrelated to this patch? http://gwt-code-reviews.appspot.com/54809/diff/38/1004#newcode694 Line 694: void compileForWebMode(String moduleName, String... userAgents) RunStyleHtmlUnit needs it, since multiple browsers might be passed in. On 2009/07/28 23:08:41, scottb wrote: > Again, this seems like a good fix, but unrelated. http://gwt-code-reviews.appspot.com/54809/diff/38/1004#newcode714 Line 714: Set<Platform> bannedSet = new HashSet<Platform>(); On 2009/07/28 23:08:41, scottb wrote: > Use an EnumSet. Done. http://gwt-code-reviews.appspot.com/54809/diff/38/1004#newcode735 Line 735: return runStyle instanceof RunStyleHtmlUnit I agree. For now, can I just put a TODO in here? On 2009/07/28 23:08:41, scottb wrote: > This should not be hard coded. The RunStyle needs to somehow specify how it > interacts with the platforms. http://gwt-code-reviews.appspot.com/54809 --~--~---------~--~----~------------~-------~--~----~ http://groups.google.com/group/Google-Web-Toolkit-Contributors -~----------~----~----~----~------~----~------~--~---