Feature #7964
open
Do not cause memory issues detection in generated environment models
Added by Evgeny Novikov about 7 years ago.
Updated about 7 years ago.
Category:
Environment models
Description
Alexander Petrenko recalled us recently that there is no much sense to find memory issues in generated environment models. There are already quite many related issues, like #7955, #7956, #7831, some of which were fixed, some still exist and result in quite bad verification results. Of course, one can fix environment model specifications and algorithms step by step, but this looks like with our numerous attempts to avoid undefined traversing through Python dictionaries. We spent very much time to fix the latter but eventually gave this up and just specify a special environment variable that forbids dictionaries randomization.
I propose to reject the issue, since the task is mostly solved by external allocating function for checking memory safety. Moreover, issues given there as related caused not by a specification incompleteness or incorrectness but by a bug in EMG algorithm for default signals generation.
This issue exactly asks for an ability to avoid detection of as many as possible memory issues in generated environment models. Likely that can be done just for incorrect memory release and memory leaks but even this does have sense. Nevertheless for developers of Klever there should be an ability to return back such issues detection, e.g. when one will want to fix EMG.
It is too abstract. We cannot propose any universal solution for this anyway. If you have any particular proposals, then it is better to open corresponding feature requests.
Let's Anton will give us such the proposal or suggest to reject this issue if this isn't reasonable.
Also available in: Atom
PDF