ARE_REAL_EQUAL is a mess

[Kurt Schwehr]

r39166 makes ARE_REAL_EQUAL even more hard to follow. This is a major maintenance hassle waiting to bite.

// Test if 2 floating point values match. Useful when comparing values
// stored as a string at some point. See #3573, #4183, #4506
#define ARE_REAL_EQUAL(dfVal1, dfVal2) \
 /* Is it FLT_MIN ? Cf #6578 */ \
 ((fabs(dfVal2) <= 3.4028234664e+38 && (float)dfVal2 == (float)1.17549435e-38 && fabs(dfVal1) <= 3.4028234664e+38) ? ((float)dfVal1 == (float)dfVal2) : \
 /* Or DBL_MIN ? */ \
  (dfVal2 == 2.2250738585072014e-308) ? (dfVal1 == dfVal2) : \
 /* General case */ \
  (dfVal1 == dfVal2 || fabs(dfVal1 - dfVal2) < 1e-10 || (dfVal2 != 0 && fabs(1 - dfVal1 / dfVal2) < 1e-10 )))

Things that should be done with this:

• An inline function
• Use std::numeric_limits<typename>::foo()
• Explain what is going on
• Have C++ tests ~/src/gdal/autotest/cpp && find . -type f | xargs grep ARE_REAL_EQUAL gives nothing
• Is this for float or double or both? Why not have a template or separate functions?
• what happens with +/- inf or NaNs?

See also:

• ​https://stackoverflow.com/questions/17333/what-is-the-most-effective-way-for-float-and-double-comparison
• ​http://en.cppreference.com/w/cpp/types/numeric_limits/epsilon

[Even Rouault]

In 39324:

Cleanup and test ARE_REAL_EQUAL() and GDALIsValueInRange() (fixes #6945, fixes #6946)