real type on Win 64 broken (?)

[dlangBugzillaToGithub]

Michael reported this on 2013-07-21T10:27:45Z

Transferred from https://issues.dlang.org/show_bug.cgi?id=10691

CC List

• Johannes Blume
• Michael
• Rainer Schuetze
• yebblies

Description

import std.stdio;

void main()
{
    writeln(real.min_normal, " ", real.max);
}

Produces every time different output:
1 run - 8.9527e-313 8.9527e-313
2 run - 1.21008e-312 1.21008e-312
3 run - 3.56e-312 3.56e-312
...

Or not recognized by ms linker?

[dlangBugzillaToGithub]

yebblies commented on 2013-11-20T22:14:56Z

Is this still broken?

[dlangBugzillaToGithub]

jblume commented on 2014-05-26T16:01:53Z

=================================
DMD v2.065 official (32-bit):
3.3621e-4932 1.18973e+4932
3.3621e-4932 1.18973e+4932
3.3621e-4932 1.18973e+4932

DMD v2.065 official (64-bit):
2.27151e-312 2.27151e-312
2.79471e-312 2.79471e-312
1.74681e-312 1.74681e-312
=================================
DMD v2.066 git MSVC2013 (32-bit):
3.3621e-4932 1.18973e+4932
3.3621e-4932 1.18973e+4932
3.3621e-4932 1.18973e+4932

DMD v2.066 git MSVC2013 (64-bit):
0 inf
0 inf
0 inf
=================================

So the random results on Win64 seem to be fixed on git.

[dlangBugzillaToGithub]

r.sagitario commented on 2014-05-26T17:37:05Z

Yeah, the output was fixed when the unittests for Win64 were enabled. Unfortunately the runtime still uses C's sprintf for printing floating point numbers, and the MS library only supports double, not real.

[thewilsonator]

As noted windows real is actually double