added win32 build system to svn

ChangeLog

@@ -14,6 +14,8 @@
 - fix a mixup with ANY hints that caused performance problems on the main
   benchmark
 - rewritten im_circle as im_draw_circle, im_circle moved to almostdeprecated
+- added IM_TYPE_RW args for inplace ops --- nip2 uses this to wrap inplace ops
+  automatically
 - special-case 3x3 makss in im_conv() for a 20% speedup
 - add IM_TYPE_RW flag for im__rw_image, helps nip2 auto-wrap inplace ops
 - im_insertplace() casts and bandalikes

TODO

@@ -1,10 +1,15 @@
 - rename most of the inplace ops as im_draw_line() etc.
 
+- how do we wrap inplace ops in C++ now? will checking the RW bit help at all?
+
 - use im__inplace_base() in more places
 
 - nip2 should be able to call RW image ops by copying them, see
   im_draw_circle_copy()
 
+
+
+
 - use D65 in cmsCreateLab4Profile() ? not sure
 
 - VImage("poop.v") should use "rd" as well? Python too?

libvips/convolution/im_conv.c

@@ -501,7 +501,7 @@ conv3x3_gen( REGION *or, void *vseq, void *a, void *b )
 	int bands = in->Bands;
 
 	Rect s;
-	int x, y, z;
+	int x, y;
 
 	/* Prepare the section of the input image we need. A little larger
 	 * than the section of the output image we are producing.

libvips/include/vips/dispatch.h

@@ -64,7 +64,9 @@ typedef void *im_object;
 
 /* These bits are ored together to make the flags in a type descriptor.
  *
- * IM_TYPE_OUTPUT: set to indicate output, otherwise input.
+ * IM_TYPE_OUTPUT: set to indicate output, otherwise input. If the IM_TYPE_RW 
+ * bit is set and IM_TYPE_OUTPUT is not set, both input and output (ie. the 
+ * operation side-effects this argument).
  *
  * IM_TYPE_ARG: Two ways of making an im_object --- with and without a 
  * command-line string to help you along. Arguments with a string are thing