On Thursday 14 November 2013 07:16:58 William S Fulton wrote:
> On 11/11/13 08:47, Geert Janssens wrote:
> > I have submitted a bug report for guile
> > (http://debbugs.gnu.org/cgi/bugreport.cgi?bug=15842). I'm not sure
> > what more I can do for the integer test.
> Thanks for looking. I'd like a green on this test on OSX, so any
> chance you could modify the test and detect for Guile < 2.0 on osx
> and then skip the failing part of the test?
Within the runtime test, I can only check for Guile < 2.0 to skip the tests. I'm not aware of a way to check for the platform I'm running on and even less so for the compiler used to compile the wrapper code. (This issue is clang only).
Is that sufficient for you if I disable the test for guile 1.8 ?
> Did you also get a chance to look at the overloading problems
> mentioned at the beginning of the thread?
I am looking at this, but it takes me a lot of time to understand how it all fits together. I finally found the number type maps and checks you were referring to in your original message.
While studying them I found they are set up quite differently from how python is doing it (which I was using as a reference). Python is using %fragments, guile is not. So I'm studying that part now to see what's the best way forward for guile. It may take some time still as I find I only progress slowly on this level. (I have more experience with gui programming than with compiler level stuff).
At least I have the runtime test for overload_numeric converted from python to guile, but it obviously still errors out because the type maps/checks are not fixed yet.