There's a problem with how coverage.py discovers all files in pyanaconda. When the source parameter is specified as module name the coverage data includes 109 files. If the source parameter is specified as full directory name then the data files contain 124 files. This is the similar to what `make ci` coverage reports produce. Unifying those gives more accurate data and makes it easier to compare test executions in the future.
Note: I had to use the getsitepackages() gimmick (proposed by dlehman) because I didn't wan't to import pyanaconda before coverage was started.
Pushed, thanks.
Closed.
anaconda-patches@lists.fedorahosted.org