I have a GUI application that I am trying to determine what is being used and what isn't. I have a number of test suites that have to be run manually to test the user interface portions. Sometimes I run the same file a couple of times with "coverage.py run file_name -a" and do different actions each time to check different interface tools. I would expect that each time I ran with the -a argument, I could only increase the code covered line count by coverage.py (at least unless new files are pulled in). However, sometimes it gives lower code coverage after an additional run - what could be causing this?
I am not editing source between runs and no new files are being pulled in as far as I can tell. I am using coverage.py version 3.5.1.