bigfloat
bigfloat copied to clipboard
Consider changing flag behaviour on NaN propagation.
MPFR sets the nan flag even when propagating NaNs; this behaviour is different from the IEEE 754 specification, where propagating NaNs doesn't set the NaN flag. We should consider changing the behaviour for BigFloat.
Just an update from gmpy2 - I'm experimenting with a context flag called quiet_nan. If True, propagating NaN does not set the NaN flag. If False, the behavior of the MPFR library is maintained. I'm undecided on the default.