Bug #122566 for Math-BigFloat: Unexpected behavior of precision(P) vs. new(,,P)

Greetings BigFloatoids, The inline example below gives the following result on my setup: Perl version v5.24.1 Math::BigFloat version 1.999715 Precision set in new(): x(P=-5): 1234.01000 x(P=-4): 1234.0100 x(P=-3): 1234.010 x(P=-2): 1234.01 x(P=-1): 1234.0 x(P=+0): 1234 x(P=+1): 1234 x(P=+2): 1230 x(P=+3): 1200 x(P=+4): 1000 x(P=+5): 0 Precision set explicitly after new(): x(P=-5): 1234.01000 x(P=-4): 1234.0100 x(P=-3): 1234.010 x(P=-2): 1234.01 x(P=-1): 1234.0 x(P=+0): 1234.01 <== My expectation: '1234' x(P=+1): 1234 x(P=+2): 1230 x(P=+3): 1200 x(P=+4): 1000 x(P=+5): 0 All is as expected, except for the P=0 case in the second loop, where the precision is set explicitly. Is there a reason why the two cases behave differently? A list of loaded modules from the above example is also attached. ======================== begin inline script ============================= #!/usr/bin/perl # # Minimal example showing unexpected (to me) difference in reported BigFloats # created via new() with a given precision, vs. creation via new() with # no precision specified, but then setting the precision afterward via # precision(). # use strict; use warnings FATAL => 'all'; use Math::BigFloat; print "Perl version $^V\n"; print "Math::BigFloat version $Math::BigFloat::VERSION\n"; # # Set precision in new(): All results as expected # printf "\n"; printf "Precision set in new():\n"; for (my $prc = -5; $prc <= 5; ++$prc) { my $x = Math::BigFloat->new("1234.01", undef, $prc); printf " x(P=%+2d): $x\n", $prc; } # # Set precision after new(): Result with P = 0 is 1234.01, expectation # is 1234. # printf "\n"; printf "Precision set explicitly after new():\n"; for (my $prc = -5; $prc <= 5; ++$prc) { my $x = Math::BigFloat->new("1234.01"); $x->precision($prc); printf " x(P=%+2d): $x\n", $prc; } ======================== end inline script =============================