Commit request: #38271, fix optimize variant of gsl

[Daisuke Takahashi]

Hi all,

Recently I overtook a maintainership of gsl port, but still waiting for the reply to my commit-right application.
So, could you please commit my last patch to gel port and close #38271? The patch was already uploaded to #38271 as Portfile.diff, and for your information, I put it at the end of this message.
It disables AVX instructions introduced at recent intel's CPUs, because assembler from cctools still does not able to assemble those instructions.

Thank you very much

Daisuke Takahashi

--- Portfile	(revision 109591)
+++ Portfile	(working copy)
@@ -84,9 +84,16 @@
     configure.cflags-append     "-ftree-vectorize -O3"
 }
 
+default_variants +gcc47
+
 platform darwin i386 {
-    if { [variant_isset optimize] && ( [variant_isset gcc43] || [variant_isset gcc44] || [variant_isset gcc45] || [variant_isset gcc46] || [variant_isset gcc47] ) } {
-        configure.cflags-append "-march=native"
+    if { [variant_isset optimize] } {
+        if { [variant_isset gcc43] } {
+            configure.cflags-append "-march=native"
+        } elseif { [variant_isset gcc44] || [variant_isset gcc45] || [variant_isset gcc46] || [variant_isset gcc47] } {
+            # Currently, gcc cannot compile AVX instructions on OSX
+            configure.cflags-append "-march=native -mno-avx"
+        }
     }
 }