[Gcjwebplugin-devel] Patch: applet viewer in a separate process

[Thomas Fitzsimmons]

Hello,

I'm checking in the attached patch.

It does the following:

- adds an applet viewer
- makes the plugin a stub that spawns the applet viewer as a separate
process

The idea is that we make the libgcj process completely separate from the
browser process and have them communicate through pipes.  This has the
following advantages over the previous JNI-based design:

- insulates the browser from libgcj crashes
- avoids the unresolved symbol problem described here:
http://mail.gnu.org/archive/html/gcjwebplugin-devel/2003-08/msg00000.html
- avoids the many threading issues that result from having two gtk main
loops in the same process (Mozilla's and the AWT's)
- allows the same viewer to be used for standalone testing of applets
and for viewing applets in the browser

Currently, the applet viewer is somewhat skeletal.  It does not parse
the applet tags from an html file.  Instead, it obtains tag information
from command line options.  It operates in one of two modes; either in
the default standalone mode or, if --xid=<socket xid> is specified, in
embedded mode.

I added a glib dependency to the plugin; I think that glib is a good
library to use for dealing with the inter-process stuff.

For the command line processing stuff, I imported the gnu.getopt package
(which I omitted from the attached patch).

I've made no attempt to design an iter-process protocol yet.  That's
something we need to discuss.

I've run some test applets with this design and even in its current
skeletal form it works better than version 0.0.2.

Tom

appletviewer.diff
Description: Text Data