Improve Coverity Scan Integration

[Gedare Bloom]

The goal of this project is to create better workflows to incorporate static analysis using Coverity Scan into the RTEMS development and release cycle.

Coverity Scan is a static analyzer that can identify various types of potential software defects. Coverity offers free use of this analyzer for free software projects. Issues identified for RTEMS are at ​https://scan.coverity.com/projects/rtems.

Coverity Scan identifies POTENTIAL issues. Some may be real bugs. Others may indicate that Coverity Scan does not have full awareness of the program life. For example, memory allocated during RTEMS initialization may appear to be leaked because it is never freed, but this is deliberate and the issue marked as such in Coverity Scan.

You can get an account on ​Coverity Scan and request access to the ​RTEMS Project to contribute.

The current process for dealing with Coverity Scan is entirely manual. See, for example, Google Code-In task instructions for Coverity. This manual process has several drawbacks. The purpose of this GSoC Project would be to simplify the process of working with Coverity, and to automate it where possible. We have identified two starting directions, and encourage interested students to brainstorm on other ways to improve the Coverity Scan integration.

Reducing False Positives
Static analysis suffers from a high number of false positives. Coverity Scan provides two methods for removing false positives. First is by creating an optional modeling file to accommodate for special cases where the data flow analysis breaks, for example due to missing functions that are linked outside the analyzed source code. A modeling file usually is used to identify functions that terminate execution, allocate memory, or free memory. Second, source code annotations in specially formatted comments can also be used to suppress warnings. This project should determine which approach to take, and design/implement a path forward.

Testing Fixes
Since security tools tend to be costly in terms of time or compute resources, they are normally run on nightly or even weekly builds rather than on every commit as done for typical continuous integration (CI). It can be tedious to merge commits to the development master and trigger a scan to determine if the issue has been fixed. Instead, we would like to develop fixes and trigger Coverity Scan as needed (subject to staying within the allowed scan rates). Coverity Scan integrates with Github and can be triggered to scan by merging new code to a specific git repository branch. As a first step, a special 'coverity' branch could be created for scanning commits that are pushed there, so that developers who are testing changes can work through the coverity branch before merging fixes into the master. Alternate solutions should be discussed with any mentors.