[PROJ] test failure with 6.3.0 geodtest

[Greg Troxel]

Even Rouault <even.rouault at spatialys.com> writes:

>> This seems to be two planimeter test cases only, and as I read the test
>> code, the rest of the planimeter cases are ok, as are all the rest of
>> the many test.
>> 
>> The system gets a clean pass from the floating-point test paranoia.
>> 
>> 
>> I assume this test passes on Linux/amd64.  I am curious what other
>> platforms it has been run on, and results.
>
> CI also tests FreeBSD 11, MSVC 2015 & 2017 and MacOSX, and all are successful 
> on the whole test suite.

Thanks, that's useful.

> I see that geodesic.c has seen non trivial changes between 6.2.1 and 6.3.0, 
> one of them being that we rely on C99 for PROJ builds. Is your setup (compiler 
> + libc) of netBSD 8 C99 compliant?

Yes, it's gcc 5.5.0 and I have declared C99 and C++11.   proj's
configure says:

  checking dependency style of gcc... gcc3
  checking for gcc option to accept ISO C99... none needed
  checking whether we are using the GNU C++ compiler... yes
  checking whether c++ accepts -g... yes
  checking dependency style of c++... gcc3
  checking whether c++ supports C++11 features by default... yes

> You might try building with -DHAVE_C99_MATH=0 to check that.

With that it passes, which helps narrow down things a lot.  I'll ask the
NetBSD math experts to look into it.

I suppose I can bisect on the various replacements the file has.