When I want to build the server.product or the swingclient.product according to http://www.globalqss.com/wiki/index.php/IDempiere/Creating_Installers_from_Eclipse I get the following message:
org.eclipse.core.runtime.CoreException: D:\NewOSGI\IDempiere\org.adempiere.server-feature\product.ant:22: Java returned: 13
The error message was first mentioned in 2012-09-10 in https://groups.google.com/forum/#!topic/idempiere/gCGs8RdTOqU
As far as I can say the error message comes and goes for unknown reasons. I have a workspace where is sometimes shows up and after trying a thousand things like workspace clean etc. it goes away. After changing something in the feature project it comes back. This needs some more research.
This issue has two parts: One part is that there is an error. The other part is that this error does not give a meaningful error message (this is why I fight for two weeks with this). The "error 13" leads by google to an eclipse issue about a wrong binary architecture. I believe this is a wrong path and the error lies somewhere in the idempiere build configuration.
To part 1:
Carlos pointed me to https://groups.google.com/forum/#!msg/idempiere/sKKA38diT2I/_4bUn7yet0sJ
It seems that many kinds of errors can lead to an "error 13". My special problem seems to come from me using Eclipse Kepler (instead of the officially proposed Indigo). So it is my fault that I use iDempiere outside of the official specifications. I will lower the Priority of this bug and declare this part of my issue as closed (until we officially update to kepler but this will mean to update our used plugins and will probably solve my bug).
Part 2:
It seems that forked java ant tasks do not write their messages to stdout.
This patch for the swingclient product ant file is an idea to give a better error message. If you feel good with this we should do the same for the server product. WDYT?
To part 1 (caused probably by new Eclipse version): For completelyness (and googling people) I give the whole error I got in the log (or on console with my patch) here. My workaround is to delete $HOME/buckminster.output/* and then retry. This seems to solve this issue (but after two weeks of trying I do not believe anything about this).