Hi G-Men, hi Bob, hi Ray, On Monday, I switched our app (I mean everyone in the team, not only on a local branch) to GWT trunk@9753 with a few patches that are still in review, in the hope that my team stop harass me with bug reports about editors and RequestFactory (my first idea was to switch to 2.2 and live with the remaining bugs, but we'll soon show the (far from finished) app to our clients and, even worse, make them test it for a couple of days, so it has to work better than what 2.2 gives us). It unfortunately turned out to have been a bad idea, as we discovered 3 major regressions introduced by the overhaul at r9666 (that "666" in the rev. number must be a sign):
1. RequestFactoryEditorDriver failing to collect path when a dotted path is given to the @Path annotation (issue 6066<http://code.google.com/p/google-web-toolkit/issues/detail?id=6066>, OMG a hidden 666 once more!) 2. EditorDelegate's getPath is wrong when the subeditor is referenced with a dotted path in a @Path annotation (issue 6076<http://code.google.com/p/google-web-toolkit/issues/detail?id=6076> ) 3. ListEditor subeditors' value is not flushed when used with a RequestFactoryEditorDriver (issue 6081<http://code.google.com/p/google-web-toolkit/issues/detail?id=6081> ) I have fixes for the first 2, but that last one is a bit trickier to fix * correctly* (I'll apply a dirty fix to our copy, running the visitor twice and living with the performance penalty; it should be OK as we're fortunately only targeting "modern browsers", our main target being Firefox 4). FWIW, here's the list of patches I also apply to make our 2.3-SNAPSHOT version: - Issue 5599<http://code.google.com/p/google-web-toolkit/issues/detail?id=5599>: HasDataEditor doesn't handle removes from the list (never calls HasData#setRowCount) - Issue 5952<http://code.google.com/p/google-web-toolkit/issues/detail?id=5952>: RequestContext.isChanged returns true just after an edit() whenever there's a List or Set - Issue 6015<http://code.google.com/p/google-web-toolkit/issues/detail?id=6015>: Editor generator can generate filenames too long for almost all filesystems - Issue 6016<http://code.google.com/p/google-web-toolkit/issues/detail?id=6016>: EditorContext generated class name should take parameterization into account - Issue 6040<http://code.google.com/p/google-web-toolkit/issues/detail?id=6040>: Editor framework does not support is / has methods (we can easily live without it, I believe we're no *yet* taking advantage of the fix) - Issue 6059<http://code.google.com/p/google-web-toolkit/issues/detail?id=6059>: ClassCastException in AbstractRequestContext.isChanged if you create()d a ValueProxy - http://gwt-code-reviews.appspot.com/1338807/ Introduces SkipInterfaceValidation annotation (I planned to remove a few of our ServiceLayerDecorator hacks and use the annotation, until the team reported me the bug I logged under issue 6081) All issues have however been verified against a fresh trunk checkout (that's when I can say I don't regret my switch to git-svn!) -- http://groups.google.com/group/Google-Web-Toolkit-Contributors
