[SWIG] mraa_init() runs twice on SWIG'd builds

[arfoll]

[SWIG] mraa_init() runs twice on SWIG'd builds

[alext-mkrs]

I've performed some research and it looks like this is because it's called first as a library CTOR and second time as a Python/JS module initialization subroutine due to line 29 in mraa.i, where we explicitly ask SWIG to call mraa_init() during module initialization.

I'm not yet that much into SWIG to understand all consequences, but after reading the basic docs and analyzing this from a common sense standpoint :) - if we have it defined as CTOR and libc does call it for us anyway, there's no point in having it in the SWIG %init section.

What do you think?

[tingleby]

Would make sense to me. The mraa.i might not have been updated when the CTOR was added.
Gave it a quick test, removed the init section around line 29 and the modules still works. (At least python)

[arfoll]

ehehe that didn't require much research, I know that. The problem is that I fixed this issue with eec08ac by adding a "nice" #ifndef. The reason for keeping it in %init is historical (ctors came after) and also because ctors get called extremely early and it's a little cleaner IMHO than the rather brutal ctor behaviour. Also we can do cool stuff if %init fails. But yes that would fix the issue.

But really the issue is why #if (defined SWIGPYTHON) || (defined SWIG) doesn't work. Which does worry me - a little. Also that statement should probably be just SWIG.

[alext-mkrs]

Ok, I see. That would've been too easy :) Indeed it does seem to ignore these there, I've just tried several variations. Looks like some deeper digging is in order here.