Hi, On Sat, 2005-01-15 at 17:31 +0100, Martin Platter wrote: > Mark Wielaard wrote: > >Does this mean CACAO now does AWT? > > > AWT support is not complete yet (specially in the released version). > Mainly there are still some threading issues to solve. But with > portable-native-sync disabled, cacao(cvs) and AWT (gnu classpath 0.12) > it works quite well. > With portable-native-sync enabled cacao/AWT doesn't work at all.
Note that there can be two issues that you can run into. One is that normally gtk+ depends on the native platform threads (usally pthreads) for locking operations. When given the --enable-portable-native-sync option you can compile version of the naitve peers that use standard runtime Monitors for all locking operations. This was written by one of the JikesRVM hackers, Steven Augert, to support the n-m thread model of JikesRVM. The portable-native-sync code is not very much tested, except together with JikesRVM, even though it should theoretically work with any runtime JNI implementation. Two is that currently the gtk+ peer code can sometimes use the wrong JNIEnv. We store a global JNIEnv* when the event queue is started and reuse that JNIEnv* almost everywhere. This is clearly wrong. We can only reuse it in the same Thread. Otherwise we have to get a new JNIEnv through the JavaVM (which we do actually store, but don't use yet except in the portable-native-sync code). Cheers, Mark
signature.asc
Description: This is a digitally signed message part
_______________________________________________ Classpath mailing list [email protected] http://lists.gnu.org/mailman/listinfo/classpath
