remove logging, fix shrink hint

vipsthumbnail works again now

TODO

@@ -1,27 +1,12 @@
-- curious
+- tune shrink's chunking to try to make it handle smalltile faster 
 
-$ time vips shrink wtc.tif[sequential] tn_wtc.jpg 100 100 --vips-progress
-vips temp-2: 2 threads, 93 x 1 tiles, groups of 768 scanlines
-vips temp-2: done in 1s          
-memory: high-water mark 58.72 MB
-real	0m1.575s
-user	0m2.164s
-sys	0m0.204s
+- add a int-only path?
 
-$ time vipsthumbnail wtc.tif --vips-progressvipsthumbnail
-temp-4: 2 threads, 128 x 128 tiles, groups of 768 scanlines
-vipsthumbnail temp-4: done in 0s          
-memory: high-water mark 44.28 MB
-real	0m2.358s
-user	0m2.216s
-sys	0m0.136s
+- compare to rightshift_size, deprecate that? should be as quick
 
-  is the real time difference just due to smalltile vs. thinstrip?
+- warn if used with float factors
 
-  can we trick vipsthumbnail into using fatstrip?
-
-- vipsthumbnail a tif with an embedded profile ... don't seem to get a profile
-  in the jpg?
+- vipsthumbnail should not remove profiles by default
 
 
 

libvips/conversion/sequential.c

@@ -34,8 +34,8 @@
  */
 
 /*
- */
 #define VIPS_DEBUG
+ */
 
 #ifdef HAVE_CONFIG_H
 #include <config.h>

libvips/iofuncs/image.c

@@ -1621,10 +1621,12 @@ vips_image_write_gen( VipsRegion *or,
 	VipsRegion *ir = (VipsRegion *) seq;
 	VipsRect *r = &or->valid;
 
+	/*
 	printf( "vips_image_write_gen: %p "
 		"left = %d, top = %d, width = %d, height = %d\n",
 		or->im,
 		r->left, r->top, r->width, r->height ); 
+	 */
 
 	/* Copy with pointers.
 	 */

libvips/iofuncs/sinkdisc.c

@@ -38,8 +38,8 @@
  */
 
 /*
- */
 #define VIPS_DEBUG
+ */
 
 #ifdef HAVE_CONFIG_H
 #include <config.h>

libvips/resample/shrink.c

@@ -62,8 +62,8 @@
  */
 
 /*
- */
 #define DEBUG
+ */
 
 #ifdef HAVE_CONFIG_H
 #include <config.h>
@@ -316,8 +316,14 @@ vips_shrink_build( VipsObject *object )
 
 	if( vips_image_copy_fields( resample->out, resample->in ) )
 		return( -1 );
+
+	/* We want to be able to work with sequential sources. THINSTRIP would
+	 * be OK, but FATSTRIP would be a disaster: thread 2 would be given a
+	 * strip some way down the output, which would be a huge distance down
+	 * the input.
+	 */
 	vips_demand_hint( resample->out, 
-		VIPS_DEMAND_STYLE_THINSTRIP, resample->in, NULL );
+		VIPS_DEMAND_STYLE_SMALLTILE, resample->in, NULL );
 
 	/* Size output. Note: we round the output width down!
 	 */