[M3commit] CVS Update: cm3

[Jay Krell]

CVSROOT:	/usr/cvs
Changes by:	jkrell at birch.	10/05/12 14:30:40

Modified files:
	cm3/m3-db/db/src/: m3makefile 
	cm3/m3-db/odbc/src/: m3makefile 
	cm3/m3-db/odbc/test/src/: m3makefile 
	cm3/m3-db/pgodbc/src/: m3makefile 
	cm3/m3-db/postgres95/src/: m3makefile 
	cm3/m3-db/postgres95/test/src/: m3makefile 
	cm3/m3-games/columns/src-x11/: m3makefile 
	cm3/m3-games/tetris/src/: m3makefile 
	cm3/m3-sys/cminstall/src/config-no-install/: Solaris.common 
	                                             Unix.common 
	                                             cm3cfg.common 
	cm3/m3-tools/cvsup/suplib/src/: m3makefile 
	cm3/m3-tools/gnuemacs/src/: m3makefile 
	cm3/m3-tools/pp/src/: m3makefile 
	cm3/m3-ui/PEX/src/: m3makefile 
	cm3/m3-ui/X11R4/src/: m3makefile 
	cm3/m3-ui/anim3D/src/: m3makefile 
	cm3/m3-ui/motif/src/: m3makefile 
	cm3/m3-ui/opengl/src/: m3makefile 
	cm3/m3-ui/ui/src/xvbt/: m3makefile 

Log message:
	move configure_assembler, configure_c_compiler, configure_linker
	from Unix.common to cm3cfg.common
	
	introduce new function: configure_system_libs
	
	The idea is:
	any code like 'if SYSTEM_LIBS contains "OPENGL"'
	(or checking SYSTEM_LIBORDER, if that is possible/done)
	
	should first call configure_system_libs
	by default configure_system_libs does nothing
	
	However on I386_SOLARIS for example, configure_system_libs
	is a theoretically slow probe to see if OPENGL is present,
	and if so, it adds OPENGL to SYSTEM_LIBS.
	
	We could just always do this check and add to SYSTEM_LIBS
	ahead of time, but it seems expensive.
	
	We could also probably do a more efficient check
	just around file existance. However the way I have written
	the probe is nicely ignorant of the precise files.
	In particular ld -lfoo -o /dev/null succeeds if libfoo is found
	and fails otherwise. That is how the probe works.
	
	We could maybe change my mind about cminstall, but no, not yet at least.
	
	To make this more efficient, we might instead offer:
	proc configure_system_lib(name)
	
	This is both a bit of a breaking change and a bit experimental
	or needs more discussion. However the level at which it
	is breaking miniscule -- it is only current relevant
	for I386_SOLARIS.
	
	I have to look again: I didn't see the Mesa stuff on 2.9/x86/Solaris.
	If it is there, we can probe alternative for it as well.
	Currently we'll just skip the opengl package on Solaris 2.9 x86.