On Saturday, June 8, 2013 1:22:30 PM UTC+2, Oliver O. wrote:
>
> Thomas, thanks for putting all the effort into this.
Thanks for the feedback!
> I've experimented a bit with 1.0-alpha-2 on a little project with a
> structure similar to your guice-rf-activities Maven archetype. What I've
> observed so far:
>
> - When used with GIN 2.0.0, on compiling *${module}.java*, the message
> "The type passed does not inherit from Ginjector - please check the
> deferred binding rules." appears for the *${module}Ginjector* class,
> followed by a stack trace.
>
> The plugin uses its own class loader while GIN's class loader delegates to
> the system class loader instead of the class loader used to load GIN
> itself. This apparently causes the same class to be loaded by different
> class loaders, undermining class identity.
>
>
>
> If the gwt-maven-plugin alpha is used with a GIN version where the fix for
> GIN issue 186 <https://code.google.com/p/google-gin/issues/detail?id=186>
> has
> been applied, it works as expected. Unfortunately, this fix is not
> available in GIN's svn repository yet.
>
>
I've opened https://github.com/tbroyer/gwt-maven-plugin/issues/5 to add an
option to fork a process.
>
> - On the first run of mvn package after mvn clean, this message
> appears, followed by a stack trace: [INFO] Unable to create new cache
> log file .../${module}/${module}-client/target/${module}
> -client-1.0-SNAPSHOT/../gwt-unitCache/gwt-unitCache-0000013F1E883EF9.
>
> Yes, I think the problem is that ${module}-client-1.0-SNAPSHOT does not
exist (yet) so ${module}-client-1.0-SNAPSHOT/../gwt-unitCache fails
(strangely, mkdirs() works, but it doesn't directly passed the absolute
path down to a native method, it first canonicalizes it)
I think this is a JVM bug, but it could be worked around in GWT (have
PersistentUnitCache call getCanonicalFile())
>
> - Successive invocations of mvn package on unchanged sources always
> compile the GWT module.
>
> This is strange because I have an integration test about this.
Could you please file a bug with a reduced repro case or some more
information?
> The third point is what currently precludes further use of the alpha-2
> version. In addition, I still need to figure out how to configure a draft
> compile for just one user agent. In the guice-rf-activities archetype
> there was a separate gwt.xml module file for that. Now it seems that module
> file names are somewhat hard coded into the plugin. However this one is not
> that important as one might introduce a code generation step for the module
> file.
>
Feel free to file an issue. Not sure how to deal with it yet. (should the
src/main/module.gwt.xml path be configurable? should things be configurable
in the pom.xml?)
> I've also read your well written analysis in the blog post In quest of
> the ultimate build
> tool<http://blog.ltgt.net/in-quest-of-the-ultimate-build-tool/>, so
> I'm curious how this might work out. For now, I would appreciate if Maven
> could be tamed to nicely support multi-module GWT builds and the new plugin
> looks promising to me.
>
That's the goal of that plugin. The CodeHaus plugin looks more like a hack
than a full-thought-out solution.
--
You received this message because you are subscribed to the Google Groups
"Google Web Toolkit" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to google-web-toolkit+unsubscr...@googlegroups.com.
To post to this group, send email to google-web-toolkit@googlegroups.com.
Visit this group at http://groups.google.com/group/google-web-toolkit?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.
