gpc_polygon_clip is bad configure test function for agg, sometimes missing

[warmerdam]

Even Rouault reports (on mapserver-dev):

Currently in configure.in, if you specifiy "--with-agg" or "--with-agg=yes" there is a check for gpc_polygon_clip to enable AGG build. AC_CHECK_LIB(agg, gpc_polygon_clip, AGG_ENABLED=-DUSE_AGG_,)

However, GPC has a non free licence ("free for non-commercial use"). See ​http://www.antigrain.com/license/index.html. So, my build of AGG doesn't include GPC.

I've removed the check for gpc_polygon_clip in configure.in and mapserver 5.0.0-beta2 builds fine.

---

I can confirm that at AGG, as packaged for free software distributions, generally does not include the GPC functionality. If we aren't using it, we should use another test for AGG in configure.

[warmerdam]

Tomas Bonfort writes:

for the time being we don't use it, and I don't plan on using it. for polygon fills there's an AGG function.

from antigrain.com: In AGG you can get the same visual result using Scanline Boolean Algebra. It operates with scanline shapes and in average works 5-10 times faster than GPC

[sdlime]

I removed the test... Configure blindly assumes that libagg is there, although I added the check for freetype as Frank had done. r6500

Steve