Re: [cp-patches] Patch: RFC: portable native sync initialization

[Thomas Fitzsimmons]

Hi,

Since I didn't get any response to this, I committed it.  I also merged
the patch on libgcj HEAD.

Tom

On Wed, 2005-04-27 at 21:59 -0400, Thomas Fitzsimmons wrote:
> Hi,
> 
> We don't use the portable native sync code in libgcj so I'm posting this
> patch here for discussion.  When the AWT parts of OpenOffice.org were
> initialized, we were seeing a GThread assertion failure:
> 
> GThread-ERROR **: GThread system may only be initialized once.
> 
> See: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=154215
> 
> It turned out that we needed to check g_thread_supported () before
> initializing the threading system.  This works fine in libgcj because it
> uses the native threading implementation, just like GTK.
> 
> Unfortunately, GThread doesn't allow overriding the threading
> implementation after it has already been initialized.  This means that
> running the AWT under OpenOffice.org on a portable native sync-using JVM
> will likely be problematic.  I'm not sure how to work around it.  Worse,
> Owen says that overriding the threading system at all, even at
> initialization-time, will be deprecated in a future release of GLib.
> 
> Does this patch look acceptable to people?  All it does is issue a
> warning when portable native sync must be disabled.
> 
> Tom
> 
> 2005-04-27  Thomas Fitzsimmons  <address@hidden>
> 
>       * jni/gtk-peer/gnu_java_awt_peer_gtk_GtkToolkit.c
>       (init_glib_threads): Check if threading system is already
>       initialized.
> 
> _______________________________________________
> Classpath-patches mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/classpath-patches