Bug 18570 – exp function does not work correctly for real in 64bit Windows

exp(real x) does not work when compiled under 64bit but does work under x86. Others have confirmed this: https://forum.dlang.org/post/[email protected] below is a simple example. import std.stdio; import std.math; unittest { writefln("ln(largest double) = %s", log(double.max)); // 709.783 writefln("e^710 = %s", exp(710.0));// inf, makes sense perhaps writefln("ln(largest real) = %s", log(real.max)); // 11356.6 real t = 710.0; writefln("e^710 = %s", exp(t)); //inf in 64bit, 2.23399e+308 for 32bit }

The exp function works correctly. The problem is writefln calling snprintf from Microsoft stdlib which does not support reals. Real parameters are downcasted to double, that's why we obtain infinity. The same will happen on 32-bit if compiled with -mscoff32. Problem: https://github.com/dlang/phobos/blob/master/std/format.d#L2532

See also: https://issues.dlang.org/show_bug.cgi?id=18326 https://issues.dlang.org/show_bug.cgi?id=9297

Very helpful Răzvan. Thanks. I hope that bug can be fixed in the future.

Thanks, seems like this is a duplicate then. *** This issue has been marked as a duplicate of issue 9297 ***