Fix a big memory leak, support a better binary format, and steps toward C++ and C API's

[xoltar]

• Create a rivet.a static library
• Beginnings of C++ and C api in that library, which is then used by rivet_console and is suitable for use in long-running systems
• Add support for msgpack-based serialization, which tends to generate files that are about half the size of those generated by boost-based serialization.
• Replace most uses of smart pointers within Arrangement and related classes, which addresses a significant memory leak that arises when using RIVET in a long-running process as opposed to doing a single calculation and exiting.

[mlwright84]

Bryn, thanks for doing all this work! The code looks good to me.

I am able to build rivet_console. However, when I try to build the RIVET viewer, I receive the following error:

In file included from dcel/barcode_template.h:26:0,
                 from computationthread.h:29,
                 from visualizationwindow.h:29,
                 from main.cpp:22:
./math/template_point.h:24:10: fatal error: msgpack.hpp: No such file or directory
 #include <msgpack.hpp>
          ^~~~~~~~~~~~~
compilation terminated.
Makefile:3425: recipe for target 'main.o' failed
make: *** [main.o] Error 1

Is make supposed to find this file?

[xoltar]

@mlwright84 Please try now, viewer should build cleanly.

[mlwright84]

Bryn, thanks for the edits. I now experienced a compile-time error for the console related to the debug statement in line 521 of dcel/arrangement.cpp. It seems that debug() << a doesn't work, where a is of type Anchor. When I changed << a to `<< a`, then I was able to compile the console, but not the viewer. I commented out this code, and then I was able to compile both the console and the viewer. After that change, everything seems to work.

I don't think we necessarily need to print the anchor in this debug statement. I'm satisfied to leave it as it is now and merge this pull request.

One small note: when running a computation from the RIVET viewer, opening the select file dialog box produces the following warning in the terminal: Gtk-Message: GtkDialog mapped without a transient parent. This is discouraged. I don't think we have to resolve this warning now, and it might only happen on some systems. I'm running Ubuntu 17.10.

[xoltar]

OK, thanks!

On Mon, Feb 5, 2018 at 12:19 PM, Matthew Wright notifications@github.com wrote:

Bryn, thanks for the edits. I now experienced a compile-time error for the console related to the debug statement in line 521 of dcel/arrangement.cpp. It seems that debug() << a doesn't work, where a is of type Anchor. When I changed << a to << a, then I was able to compile the console, but not the viewer. I commented out this code, and then I was able to compile both the console and the viewer. After that change, everything seems to work.

I don't think we necessarily need to print the anchor in this debug statement. I'm satisfied to leave it as it is now and merge this pull request.

One small note: when running a computation from the RIVET viewer, opening the select file dialog box produces the following warning in the terminal: Gtk-Message: GtkDialog mapped without a transient parent. This is discouraged. I don't think we have to resolve this warning now, and it might only happen on some systems. I'm running Ubuntu 17.10.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/rivetTDA/rivet/pull/106#issuecomment-363208998, or mute the thread https://github.com/notifications/unsubscribe-auth/ABK13LJHBiDV360cdtNHCBiQJD97cxrIks5tR2JvgaJpZM4RJbu2 .