Sumit Garg

Looking at reason behind introduction of `burn_stack()` api [1], it seems like in case of a lot of stack variables, clearing every variable inside function makes it look bit messier...

@ksherlock Performance wise approach suggested by you looks better than "Approach 2". The only thing that doesn't looks good is a `define` for every variable which makes function look a...

Any more preferences among above 3 approaches (Approach 3: suggested by @ksherlock)? Or any further suggestions to address this issue?

> I already thought about an approach 4 where we leave the burn_stack() approach but increase the buffer size, so the overshooting isn't that severe. What should be the appropriate...

Thanks everyone for your feedback. Will use "Approach 2" to create a fix patch.

It seems likely that we can't get agreement on one of above discussed approaches. So how about following simple change to use variable length array instead of recursive call in...

@rofl0r Agree but with a minimal change like VLA, we could improve the `burn_stack()` API atleast. It was just an effort to make current code workable rather than making developer's...

@arfoll Thanks for these patches. But I am facing following cmake configuration errors after applying this patch: ``` CMake Error at src/python/python2/CMakeLists.txt:21 (add_dependencies): Cannot add target-level dependencies to non-existent target...

``` root@linaro-developer:~/mraa/build# make help The following are some of the valid targets for this Makefile: ... all (the default if no target is provided) ... clean ... depend ... install/local...

It seems that display format of version info changed from sphinx v1.3.5 to v1.7.6 as follows: ``` $ sphinx-build --version In case of v1.3.5: "Sphinx (sphinx-build) 1.3.5" In case of...