Boost C++ Libraries: Ticket #12512: cpp_bin_float: convert_to for numbers with absolute values smaller than 2**(INT_MIN+52) returns infinity instead of 0.

attachment set

Michael Shatz — Mon, 10 Oct 2016 19:52:21 GMT

attachment → cnv_under_test.cpp

demonstrates cpp_bin_float convert_to<> bug

Michael Shatz — Tue, 11 Oct 2016 10:07:29 GMT

Your posting engine changed my post above in unexpected way. I think, it would be beneficiary for non-sophisticated people like me, if by default you engine will have WikiFormatting turned off. An option to turn it on can be provided for those of us who know what they are doing.

I meant to post:

Applying convert_to<double>() to cpp_bin_float object with absolute values in range (0..2**(INT_MIN+52)] returns infinity.
Similar bug exists with convert_to<float>().
Due to limits on exponent of cpp_bin_float the bug can be seen only when class Exponent = int64_t or wider.

status, milestone changed; resolution set

John Maddock — Mon, 24 Oct 2016 17:49:01 GMT

status new → closed
resolution → fixed
milestone To Be Determined → Boost 1.63.0

Confirmed, fixed in https://github.com/boostorg/multiprecision/commit/ef0ef9b515aa63c80b71863dc0992261acdd82b9.