3.5.0-dev compilation errors on MinGW

[esseffe]

while attempting to build 3.5.0-dev on Windows using MinGW I've discovered that the compilation fails due to multiple definitions of inlined symbols: you'll find the full error list into the attached file.

./configure --disable-inline

only after disabling inlined symbols I was finally able to successfully complete my build. I've tested both gcc 4.8.1 (32 bit) and gcc-tdm 4.9.2 (64 bit) and the errors were exactly the same in both cases.

it's probably useful reporting that 3.4.2 sources nicely compile without any error using the same identical MinGW compilers, so it definitely seems to be something specifically caused by 3.5.0 itself.

[robe]

Sandro (esseffe),

If you can provide a patch, that would be swell and I'll test it on my mingw install, because you are more of an expert debugging this stuff than I am, and the other Sandro won't touch anything that involves mingw64 or smell of windows with a 10-foot pole.

FWIW: Like I said on mailing list, I use CMake for building GEOS under Mingw64 (for both 32-bit and 64-bit) and don'tuse the non-w64 chain anymore.

Thanks, Regina

[robe]

I wonder if this has something to do with this r3971

I'll push back to 3.5 if I can figure it out.

[robe]

Okay if I revert back the above to what it was before the r3971, I can compile without inline errors, though I get all these failures on make check which I don't when running with cmake and cmake is way faster.

I can't figure out what to put in to detect MINGW64, I thought _MINGW64_MAJOR would do it as I have in PostGIS but couldn't get that to work.

Commits r4075 (I know that one I screwed up but trying to put the &&!MINGW64 did not work) I ended up just putting back the DLL_IMPORT check but leaving the GEOS_EXPORT.. at r4076

[robe]

strk, pramsey, esseffe, mloskot please check to make sure I didn't screw up compile for someone else.

[robe]

Still works for Mac according to pramsey, travis is also happy. I will be happy if we can release soon.