I started work on a RAM display window for Nonpareil's GUI debugger. The first attempt resulted in a window taller than the screen; I had to learn how to use a GTK viewport to get scrolling capability. Now the window starts out way to small, but at least is resizable. I'll have to figure out how to set the initial size to something reasonable.
Like the register display window, it isn't yet fully functional. Although they both use editable entry boxes for the display, if you edit them the change is not actually written back to the register or memory. And they don't update automatically, so you have to close the window and reopen it to see simulation changes. Still, it's been somewhat useful for debugging.
I'm not happy with the use of the standard GTK+ entry widget. I really want a widget that displays a number using a fixed base and digit count, in a monospaced font (or at least with the characters in fixed column positions), so that the digits in all the registers or RAM locations line up vertically, and there isn't extra empty space at the right side of the widget. Colored highlights for the P and Q pointer positions would also be useful. I'll have to write such a widget myself, and I probably won't get around to it for a while.
Google revealed that Adam Sampson has built a GARstow package for Nonpareil, and was frustrated with SCons. I discussed it with him by email, and all of his problems were really due to inadequacies in the way I used SCons, rather than with SCons per se. For instance, my SConstruct/SConscript files supported a "prefix" configuration option, but not "bindir" and "libdir" for more flexible filesystem layout, nor "DESTDIR" for a virtual filesystem root for installation packaging. Also, doing "scons install" was putting ".sconsign" files into the installation directories, which was resulting in file conflicts between various packages that use SCons.
I expect the same problems would affect almost any packaging system, including my own anticpated effort to offer RPMs for Fedora Core 3 and 4, so I've fixed them today for release 0.66. I added "bindir" and "libdir", as well as "destdir" (note lower case for consistency with other Nonpareil configuration options).
Also, by putting an "SConsignFile()" directive in the SConscript, it will put all of the signature information it uses for dependency analysis and tracking into a single ".sconsign.dblite" database file in the top directory, so it no longer puts .sconsign files into the install directories.
I added caching of the configuration options in "nonpareil.conf" to replace my previous "local.py" configuration file. And a "Help" directive so that "scons -h" will show all of the configuration options with their default and current values. To revert all options to the defaults, you can simply delete the "nonpareil.conf" file from the top directory.
I updated the INSTALL file as well as the credits in the README, added the people from the credits to the About dialog box, and moved the code for the About dialog box into separate about.c and about.h source files.
Thomas Olesen has sent me his patches to get the HP-41CX running, based on earlier modifications to nsim by Chris Roccati. I should have that integrated for the next release.
I sent Maciej Bartosiak email about the possibility of using POSIX threads in place of the glib thread support on MacOS X, and it turns out he'd already done that. He sent me his modified version of proc.c.