facebook / sparta Goto Github PK

Hello spartans (Salut Arnaud!),

I think there is a bug in the PatriciaTreeMap::leq() implementation.

If both trees are leaves, you should check if the keys are the same before comparing the values:

https://github.com/facebookincubator/SPARTA/blob/b1d453e871941b095e0297b6f8065e1230729309/include/PatriciaTreeMap.h#L500

Best regards

Hi Team,

Thank you for releasing SPARTA. I am trying to explore different functionalities and APIs of SPARTA.However I am not able to find any document which explains the usage of SPARTA. Please let me know if there is any detailed document and examples to use SPARTA for different applications.
@int3 @facebook-github-bot , Any help is highly appreciated.
Thank you.
Ashis

Hi,
I followed installation steps and got the following error (Ubuntu 18.04):

$ mkdir build-cmake
$ cd build-cmake
$ cmake ..
$ cmake --build . 
make: *** No targets specified and no makefile found.  Stop.

and indeed there is no Makefile inside SPARTA/build-cmake.
What am I doing wrong ? Thanks !

Hello together,

first of all: Thank you for releasing SPARTA on Github. As someone highly interested in static analysis and abstract interpretation I'm really excited about this. So again: Thank you.

I've got a question regarding your use of CRTP in the definition of AbstractDomain and AbstractValue. Nevertheless these classes are designed to use CRTP, they contain purely virtual functions.

Regarding libraries using CRTP you often see a function like this

Derived& derived() { return *static_cast<Derived*>(this); }

and a const version of that. A purely virtual function virtual void foo() = 0; can then be replaced by

inline void foo() { return derived().foo(); }

A method call to foo will then be completely statically dispatched. This would maybe decrease the binary size (since vtables are smaller) and could increase the performance (due to static dispatching). We use this approach at the high performance linear algebra library Eigen and I guess it works quite well.

I see the following downsides:

1. Code gets more complicated and looks less clean
2. No inbuilt compile-time check if a subclass implements a method (a call to a former purely virtual function will seg. fault at runtime). However, since SPARTA already depends on Boost, we could use Boost TTI to add checks in the (still virtual) destructor.
3. No inbuilt override (and I have no idea how to mimic one)

I guess performance is crucial if you want a static analysis tool to be widely adopted and therefore want to advertise the changes suggested above.

If you are interested in a pull request addressing the ideas above, I would prepare something.

Cheers,
David