Is anybody working on the C backend?

[fHachenberg]

I think, creating a shallow C interface still would be of great benefit because many languages already have binding generators for C.

The most recent code for this backend seems to be in the branch gsoc2012-c
https://github.com/swig/swig/tree/gsoc2012-c

Has anyone looked into this? Is it beyond repair?

Here's the description and summary from gsoc 2012
https://opensource.googleblog.com/2008/11/swigs-first-google-summer-of-code.html
https://opensource.googleblog.com/2012/11/swig-swings-into-another-successful.html
http://swig-devel.narkive.com/jqdNrIxV/gsoc-c-backend-resume

[vadz]

The most recent (but still not really functional) version of this is in my fork, see this recent thread

[fHachenberg]

I would like to help. Where do I start?

[vadz]

Pull requests fixing any of the existing test failures would be surely welcome. Perhaps the best would be to run the current code with your own header and fix the problems affecting your project (and maybe discover new ones :-)

Personally, my next goal is adding shared_ptr support as I absolutely need this for my own use, but, unfortunately, I'm not sure when I will have time to finish this...

Thanks in advance for your help!

[fHachenberg]

I started to look into the test failures.
In some test cases I get a warning for an undefined SWIG_exit. Should the runme-files include swigmod.h or should the generated wrap-file include swigmod.h or should SWIG_exit just be replaced by return? I didn't find any documentation about SWIG_exit

[fHachenberg]

Is there a design document available somewhere describing basic decisions regarding the C backend? For instance I wonder how are C++ exception mapped? Do all wrapper functions use their return value for error codes?

[ojwb]

There's a SWIG_exit() function defined in Lib/c/c.swg which just calls exit() - I'm not sure why that redirection was thought useful.

In the source of SWIG itself, there's a SWIG_exit() function which is used instead of exit() in the SWIG source code - this has an optional infinite loop, presumably for debugging purpose, though this smacks of someone not knowing how to use their debugger properly:

void SWIG_exit(int exit_code) {
  while (freeze) {
  }
  exit(exit_code);
}

I don't know if there's a design document, but I'd expect the wrappers to return the return value of the wrapped function by default - exception handling would then be handled via %exception.

[wsfulton]

The only documentation I can think of is the Doc/Manual/C.html file in the branch. If you find anything wrong in here, please feel free to correct it.

[vadz]

You can see the exception handling in action in Examples/c/exception. In short, it uses a global (ideally should per-thread) object to store the thrown exception, which can then be retrieved and examined from the C code.

This is definitely not ideal from C point of view, but this should be good enough for my ultimate goal of C++ wrapper for C++ API as such API could check for exception after every C function call and just rethrow it immediately. And other alternatives are not really better anyhow (but let's discuss them on the mailing list if you're interested).

[wsfulton]

Please try out the proposed pull request to add in C support in #2086.

[ojwb]

I'm going to close this as the question "Is anybody working on the C backend?" has been answered, and there's an open PR to track progress.