After some lengthy experimentation with test programs (which is something I love to do, honestly) (even if the slow iteration cycle on Android is a bit irrtating), the root cause to my problem seems pretty obvious. The concept of field ID in "normal" JVMs and in Android's Dalvik appears to be quite different.
In normal JVMs, a field id is a small integer, in no way unique. If two classes both have a field with a certain name and no other fields, the id of that field is the same for both classes. The field id of Context in com.sun.star.uno.Exception and in com.sun.star.uno.RuntimeException are the same. In Dalvik, field ids are basically pointers to some internal structure. And they indeed seem to be unique for each field in each class. So it is quite obvious that what the code apparently tries to do doesn't work. In the IDL and C++, css::uno::Exception has a Context field. css::uno::RuntimeException is a subclass of css::uno::Exception with no fields of its own, it inherits Exception's Context field. But the corresponding Java UNO types, in ridl.jar, have a different relationship. com.sun.star.uno.RuntimeException extends java.lang.RuntimeException and has a Context field, and com.sun.star.uno.Exception extends java.lang.Exception and has its own Context field. Hmm. OK, I think I get it, Java's RuntimeException (and the subclasses of that) has significantly different semantics than that of its Exception (and subclasses), unchecked vs. checked. The intent is that the UNO exceptions be similarly divided when viewed on the Java side. Presumably that can not be changed. --tml _______________________________________________ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
