Alas, I just discovered that the gamma() function in OS X computes the actual Gamma function, not the natural log, like the one I grew up with does (and the one on other Unix systems). This means that the version of AFNI you just compiled will not compute the 'p' value of t or correlation statistics correctly.
I have modified the AFNI source code distribution (and OS X Makefiles) along the lines of what the earlier discussion indicated. I have also fixed the use of gamma() to lgamma(), which seems to be the more portable way to get ln(Gamma(x)).
I also think I fixed the problem with 3dDeconvolve_f, but can't be sure since I don't have the problem here that you had. You can at least give it a try.
If you download and use Makefile.macosx_10.3 or Makefile.macosx_10.3_G5, you should have less trouble now. (You'll probably still need to do something to get the Motif library you compiled to link in.) I won't have a Makefile.macosx_10.4 until I get Tiger myself.