The code used to control the precision of floating point numbers is too
restrictive, the limit should be defined more like this
(2 + std::numeric_limits<float>::digits * 3010/10000)
see
http://www2.open-std.org/JTC1/SC22/WG21/docs/papers/2005/n1822.pdf
not sure if the default should be changed, for consistency I would
Original issue reported on code.google.com by Kevin.J....@gmail.com on 30 Jan 2015 at 1:57
Original issue reported on code.google.com by
Kevin.J....@gmail.com
on 30 Jan 2015 at 1:57Attachments: