Code analysis with Valgrind and Infer

[mindhells]

I'm working on this branch. There I've added Valgrind and Infer tools for the C++ examples.

Examples readme has also been updated there explaining how to run the examples using Valgrind.
No memory leaks detected so far in the C++ model... and that's the reason me opening this issue: is that true or am I doing something wrong?

[edyoung]

Cool! You could try introducing av Leah d deliberately and see if it is reported...

[mindhells]

Good point. I guess I assumed there should be some, maybe because I underestimated the current implementation.

I know this analysis won't prove much about Gnofract4d running in GUI mode. For in that case, there's "objects" being reused or disposed depending on the use case (changing the formula o zooming in/out), and to find if there's leaking there I suppose you'll need python specific tools like Pympler.

I still find tricky, when creating Python extensions, to deal with reference counting. I wonder if there's a manageable way to find leaking problems there.

[edyoung]

Re: leaking problems, I had a crappy manual way to deal with that. If you enable #DEBUG_ALLOCATION=1 you get some debug spew which logs when each python-referenced object is created or destroyed. You run the program, record the trace, then sort the output. If there is a line with 0x12345 : FOO : CTOR not followed by a 0x12345 : FOO : DTOR then you leaked a FOO. Crude, but found a number of bugs that way.

[mindhells]

No time to try it out yet, but I'll drop here this interesting piece of info, just in case someone runs into this issue: https://pythonextensionpatterns.readthedocs.io/en/latest/debugging/valgrind.html#building-python-for-valgrind-label