In-place fills have a risk of being incomplete if there is any chance
that the fillval is already present in src. This conservatively
adds a warning. There may be cases where this warning isn't needed, if
the user has special knowledge that the values are actually distinct.
But if so, it doesn't seem unreasonable to encourage the user to
exclude them from f.
This deletes one previous test, because the fact that it works is very
particular (it depends on starting at the one-and-only connected entry
with value fillvalue). The new warning gets triggered in that
former test, and that's probably a good thing.
In-place fills have a risk of being incomplete if there is any chance that the
fillval
is already present insrc
. This conservatively adds a warning. There may be cases where this warning isn't needed, if the user has special knowledge that the values are actually distinct. But if so, it doesn't seem unreasonable to encourage the user to exclude them fromf
.This deletes one previous test, because the fact that it works is very particular (it depends on starting at the one-and-only connected entry with value
fillvalue
). The new warning gets triggered in that former test, and that's probably a good thing.