Weird behavior 1_000_000_000.12345.to_d in bigdecimal 3.0.1

[kamipo]

I've found weird behavior in bigdecimal 3.0.1 when I was testing rails/rails#42125.

Expected behavior

1_000_000_000.12345.to_d #=> 0.100000000012345e10
1.000_000_000_12345.to_d #=> 0.100000000012345e1

Actual behavior

1_000_000_000.12345.to_d #=> 0.1e10
1.000_000_000_12345.to_d #=> 0.1e1

	2.0.2	3.0.0	3.0.1
1_000_000_000.12345.to_d	0.100000000012345e10	0.100000000012345e10	0.1e10 (!?)
1_000_000_0001.2345.to_d	0.100000000012345e11	0.100000000012345e11	0.100000000012345e11
1_000_000_00012.345.to_d	0.100000000012345e12	0.100000000012345e12	0.100000000012345e12
1_000_000_000123.45.to_d	0.100000000012345e13	0.100000000012345e13	0.100000000012345e13
1_000_000_0001234.5.to_d	0.100000000012345e14	0.100000000012345e14	0.100000000012345e14

[mrkn]

Thank you for the report. I couldn't remember the bugs including this are remaining in the master branch before releasing 3.0.1. 🤦