enabling -O2 with gcc 2.96 and later

[Phil Edwards]

> > My first approach was wonderfully freakin' braindead.
>
> No offense intended, but so is this second attempt, IMHO :-)

You didn't see my first attempt.  :-)


> AC_EGREP_CPP([ok], [
> #if (__GNUC__ == 2 && __GNUC_MINOR__ > 95) || __GNUC__ > 2
> ok
> #fi
> ], [OPTLEVEL='-O2'], [OPTLEVEL=])
> AC_SUBST(OPTLEVEL)

Yes, I like that much better.  (The last time I tried using EGREP_CPP,
bad things happened.  But this works.)

Since the optimization level was set in AM_CXXFLAGS before, that's where
I'd left it.  It's been moved to CXXFLAGS now.

Only one other issue before I check this in, barring protests:  Theodore
Papadopoulo suggested that

   Maybe you want to add -fstrict-aliasing for 2.96 as it has 
   been (temporarily?) invalidated....

I'm not qualified to speak on how aliasing issues affects the contents
of the src directory (since the Makefiles in, say, the math subdir,
aren't being changed).


(If you reply to the list, please don't cc another copy to me.  Thanks!)
Phil