static analysis

[jbmouret]

to try: http://clang-analyzer.llvm.org/

[costashatz]

I performed a static analysis and I found the following results:

Analyzing them a bit, we can see that we do not really have any trouble in our code except for one minor dead assignment in sys.hpp (which I will remove). There are some warnings/bugs in the ehvi and the hypervol code though. I looked a little bit deeper and the most dangerous one is the memory leak one (as I have noted in the image).

I will fix these 2 issues. The other ones are not that obvious on how to fix, since it is not our code.

Do you have any suggestions on how we can automate this static analysis procedure?

[jbmouret]

Great! Indeed, the issues are not from our code and this would not be easy
to debug. As these are still experimental components, I suggest we do not
anything for now.

I think we should add this to a weekly CI on our cluster (with the full
benchmark and the extensive tests). Ideally, we would post these results on
a mailing list and/or on github.

-- JBM

On Sat, Jul 23, 2016 at 8:31 AM Konstantinos Chatzilygeroudis <
[email protected]> wrote:

I performed a static analysis and I found the following results:

[image: static_analysis]
https://cloud.githubusercontent.com/assets/1283922/17078383/53ebe00a-50f2-11e6-93b6-b43af498e76f.png

Analyzing them a bit, we can see that we do not really have any trouble in
our code except for one minor dead assignment in sys.hpp (which I will
remove). There are some warnings/bugs in the ehvi and the hypervol code
though. I looked a little bit deeper and the most dangerous one is the
memory leak one (as I have noted in the image).

I will fix these 2 issues. The other ones are not that obvious on how to
fix, since it is not our code.

Do you have any suggestions on how we can automate this static analysis
procedure?

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#31 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/AFu8v4gDzpWMt-COODmY5S0QFQpAzGI1ks5qYiWwgaJpZM4GZwRZ
.

[costashatz]

Great! Indeed, the issues are not from our code and this would not be easy
to debug. As these are still experimental components, I suggest we do not
anything for now.

Agreed.

I think we should add this to a weekly CI on our cluster (with the full
benchmark and the extensive tests). Ideally, we would post these results on
a mailing list and/or on github.

OK. I will send you the commands that are needed to run the static analyzer. Although, I would say that the frequency should be bi-weekly not weekly.