Bug #122575 for Math-BigFloat: In some situations, new() ignores caller-specified P

This seemingly straighforward BigFloat creation $y = Math::BigFloat->new($x, undef, P); # P is an arbitrary integer seems to ignore P if $x is an extant BigFloat that already has a P value. The resulting P of $y is always the same as the P of $x. Minimal demo script below. List of loaded modules (via perldebug 'M') attached. This was surprising to me. Didn't see anything in the doc that might account for this behavior, but perhaps I missed it. Is it intentional? If so, please explain rationale. Thanks! ======================== begin inline script ============================= #!/usr/bin/perl # # Minimal example showing that BigFloat->new() can sometimes create # variables that are slightly used or "pre-owned". # use strict; use warnings FATAL => 'all'; use Math::BigFloat; print "Perl version $^V\n"; print "Math::BigFloat version $Math::BigFloat::VERSION\n"; # # Create $x using new(), specifying P = -10. # my $x = Math::BigFloat->new("1234.56789", undef, -10); print "P(x): " . ($x->precision() // 'undef') . "\n"; # # Create $y using new(), specifying $x as the value, and specifying P = -23. # Result is unexpected (to me): No matter what value is supplied for P when # creating $y, the reported precision of $y is the same as the precision of $x. # my $y = Math::BigFloat->new($x, undef, -23); print "P(y): " . ($y->precision() // 'undef') . "\n"; ======================== end inline script =============================