[GRASS-dev] reorder tcltk/opengl checks in configure?

[William Kyngesburye]

On May 14, 2009, at 7:50 PM, Glynn Clements wrote:

>
> William Kyngesburye wrote:
>
>> http://trac.osgeo.org/grass/ticket/598
>>
>> configure needs to temporarily strip out 64bit arch flags on OSX for
>> the TclTk Aqua tests.  To do this, based on the OpenGL setting
>> selected (TclTk Aqua is dependent on OpenGL setting = aqua), the
>> opengl test must happen before the TclTk tests.
>>
>> I don't see anything dependent on TclTk that would be affected if the
>> tcltk tests are moved to after the opengl tests.
>>
>> Comments from the configure/make guru?
>
> Will NVIZ cope with a 32-bit Tcl/Tk but 64-bit OpenGL?
>
> When it comes to linking NVIZ, the command will include both the
> Tcl/Tk linker flags and the OpenGL linker flags. There's no point in
> configure testing whether linking a test program against Tcl/Tk works
> without certain linker flags if NVIZ itself is going to be linked with
> those flags.
>
The system frameworks (including OpenGL) on OSX 10.5 are all quad- 
architecture - ppc/intel, 32+64bit.  Only the deprecated Carbon stuff  
that wxPython and Tk rely on are 32bit-only.  If the system was all  
only 64bit, we'd have a 64bit wxPython and 64bit TclTk by now.

So, when I configure for a quad-arch GRASS (adding arch flags to  
CFLAGS and LDFLAGS), it will test the TclTk linking with the same, and  
fail to link because 64bit arches are missing.

I plan to add a $(subst) command to the nviz makefile (and a couple  
others that link Tk libraries) to strip out the 64bit flags on OSX  
when OpenGL aqua is used.  I have that worked out and tested (same  
thing I used in the wx nviz and vdigit makefiles before the switch to  
setup.py), it just needs the configure tk tests to pass.


-----
William Kyngesburye <kyngchaos*at*kyngchaos*dot*com>
http://www.kyngchaos.com/

All generalizations are dangerous, even this one.