python mapscript setup.py doesn't handle framework link options

[kyngchaos@…]

If a Mac OS X framework or framework path option are passed to setup.py, they are dropped because 
the setup script strips unknown options.  This patch did the trick for me:

--- setup.py-orig	2006-05-31 09:48:31.000000000 -0500
+++ setup.py	2006-12-22 20:57:35.000000000 -0600
@@ -57,18 +57,27 @@
 
 libs = []
 extras = []
+ex_next = False
 
 for x in lib_opts:
-    if x[:2] == '-l':
+    if ex_next:
+        extras.append(x)
+        ex_next = False
+    elif x[:2] == '-l':
         libs.append( x[2:] )
-    if x[-4:] == '.lib' or x[-4:] == '.LIB':
+    elif x[-4:] == '.lib' or x[-4:] == '.LIB':
 	    dir, lib = os.path.split(x)
 	    libs.append( lib[:-4] )
 	    if len(dir) > 0:
 	        lib_dirs.append( dir )
-    if x[-2:] == '.a':
+    elif x[-2:] == '.a':
+        extras.append(x)
+    elif x[:10] == '-framework':
         extras.append(x)
-        
+        ex_next = True
+    elif x[:2] == '-F':
+        extras.append(x)
+
 libs = unique(libs)
 
 # if we're msvc, just link against the stub lib


Frameworks use the form: -framework foo, so are two words, thus the ex_next tracking.  Framework 
path flags are like library path flags, but with -F.  Distutils doesn't appear to know about frameworks, 
so I dump them all into extras.

[dmorissette]

Reassigned bug to owner of Python-MapScript component

[hobu]

William,

I applied this in the head branch.  Please confirm.

Howard

[kyngchaos@…]

looks like part of it didn't get applied:

+    if ex_next:
+        extras.append(x)
+        ex_next = False

ex_next = False is missing.  This is needed to reset the -framework flag after grabbing the value following 
it.  Otherwise, *everything* after a framework flag will dumped into extras.

[hobu]

added

[kyngchaos@…]

Looks good to me.

[kyngchaos@…]

I see that this patch didn't make it into the 4.10.1 release.  Maybe because it didn't get closed?

[dmorissette]

The patch was applied only in the development version (V4.99, CVS HEAD) and has
not been backported to the 4.10 branch. It is up to Howard to decide whether it
is safe to backport this to the 4.10 branch to be included in a future 4.10.2
release.

[hobu]

I just spaced this one out as far as backporting.  William, I have applied it in
the 4.10 branch.  Can you confirm it does what it is supposed to now?

Howard

[kyngchaos@…]

After patching my 4.10.1 source, it still works as desired.  It looks right in CVS (rev 1.19.2.1).