Line number count in <python:trace> test coverage report inaccurate

[cmlenz]

The counting of line numbers by the <python:trace> report is inaccurate if there are source files that don't have any coverage.

This is due to differences in how trace.py and bitten.util.loc​ count lines: The latter counts every non-commented, non-blank line that contains Python code, excluding “docstrings”. In particular, when statements spanning multiple lines are encountered, each of those lines is counted.

trace.py only counts lines that actually contain statements (which exludes for example lines that only contain else:), and a statement spanning multiple lines is count only as a single line.

The result is that the line counts for files without any coverage are in general higher than for those that have coverage. Thus, the graph displaying the percentage of total coverage will show a lower coverage than it should (as long as there are files without coverage).

When parsing trace.py output files to determine the number of lines, Bitten should also take the results of the loc.count() into account, and map the line hits onto the line numbers found by loc.count().

[cmlenz]

Should be fixed in [303].

[cmlenz]

A couple of regressions here:

• The line count for covered files now includes commented/blank lines, which it obviously shouldn't
• The percentage is completely wrong.

[cmlenz]

Fixed in [304] and [305].