I am increasing the threshold even more for when
codecov complains. Since the tool reports so many
false negatives, we should probably change our philosophy
on code coverage so that this tool only reports
catastrophic commits (like .e.g. a commit that prevents
running a whole test suite) and let our more specific
tools continue to catch things like regressions of files
from 100% coverage.
We want a little bit of wiggle room for code coverage.
Some legitimate commits that remove code from the system
appear to reduce overall code coverage.
The `patch` option for code coverage only reports on lines that
are part of the patch, which tends to be misleading for things
like refactoring commits that may actually increase overall
code coverage for the project.
Steve Howell
Steve Howell
Umair Khan