coverage:
status:
project:
default:
target: auto
# Tolerate small, cosmetic dips in total coverage. Refactors that move
# code around can shift codecov's line attribution and nudge the project
# ratio down by a few hundredths of a percent even when the new code is
# fully tested. Without a threshold, codecov's default 0% gate fails the
# project status on these dips. 0.5% absorbs that noise while still
# catching meaningful regressions.
threshold: 0.5%
patch:
default:
target: 90%