You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently `test-float-parse` is mostly random tests which take a fixed
seed, meaning the same set of values get tested for each invocation.
This isn't ideal because we don't get the true benefit of randomness,
which is to have better coverage over time.
Improve this by using a randomly generated seed, which can also be set
via env. The seed is printed at the beginning of each run so it is easy
to reproduce failures using the same test set, if needed.
Typically it isn't great to have fuzzing randomness in tests that get
run in CI because it can lead to spurious failures. However, this is a
test for which failure should never happen becasue the algorithms are
reasonably well proven, so if one does occur it represents a very
unexpected bug that needs to be addressed. The error message is updated
to strongly recommend reporting before retrying and to include details
on how to reproduce.
0 commit comments