Tcl Source Code

Tested on various windows builds only.

isunordered unexpectedly returns 1 for a range of negative values when they are expressed in integer form. 

I'm not sure of the usecase for negative values here, but it doesn't seem to match the documentation as they should be promotable to floats, and as neither argument is NaN - the 1 returns are unexpected.

%tcl::mathfunc::isunordered 1 -1.0
0
%tcl::mathfunc::isunordered 1 -1
1

This occurs for negative values expressed in integer form up to negative 2**52
for numbers below that - we again get a zero result.

%tcl::mathfunc::isunordered 1 -4503599627370495
1
%tcl::mathfunc::isunordered 1 -4503599627370496
0

Indeed, values of FP_* are compiler/implementation specific, e. g. for MSVC FP_NORMAL is equal -1, thus rewritten without FP_CLS_ERROR, because error-prone (whatever corrected value could still cross them later by some other platforms/compilers).
Fixed now in [3059fadfe6daf175].

There are still test failures with MSVC on Windows

Indeed, tcl::mathfunc::isunordered has mistakenly considered non-double numbers (e.g. int) as double.
Fixed for 8.7+, thanks for the report.