tiff pyr write fix

ChangeLog

@@ -1,3 +1,6 @@
+12/5/10 started 7.22.1
+- fix a problem with tiff pyramid write and >1cpu, thanks Ruven
+
 12/5/10 started 7.22.0
 - bump and rename
 - vipsthumbnail has a manualpage and sharpens correctly

TODO

@@ -1,3 +1,13 @@
+- write_tif_tile() single-threads, but will be called from many
+  workers ... so the tile cache fails with region ownership problems
+
+  either make vips_sink_tile() only call the write function from one thread,
+  or swap region ownership about a lot
+
+- drop "set" suffix renames im_copy_set as copy_ argh
+
+  I guess this causes nip2 problems as well? or we call the full name I guess
+
 - lcms2 needs testing
 
 - tools subdirs are now pretty stupid :-( just have a single dir
@@ -8,7 +18,6 @@
 		python setup.py build
 		python setup.py install
 
-now we get:
 
 >>> from vipsCC import *
 Traceback (most recent call last):

configure.in

@@ -6,7 +6,7 @@ AC_CONFIG_MACRO_DIR(m4)
 # user-visible library versioning
 m4_define([vips_major_version], [7])
 m4_define([vips_minor_version], [22])
-m4_define([vips_micro_version], [0])
+m4_define([vips_micro_version], [1])
 m4_define([vips_version],
           [vips_major_version.vips_minor_version.vips_micro_version])
 
@@ -29,7 +29,7 @@ PACKAGE=vips
 #   interface changes not backwards compatible?: reset age to 0
 
 LIBRARY_CURRENT=29
-LIBRARY_REVISION=1
+LIBRARY_REVISION=2
 LIBRARY_AGE=14
 
 AM_INIT_AUTOMAKE($PACKAGE,$VERSION)

libvips/format/im_vips2tiff.c

@@ -112,6 +112,9 @@
  * 	- option to turn on bigtiff output
  * 16/4/10
  * 	- use vips_sink_*() instead of threadgroup and friends
+ * 22/6/10
+ * 	- make no-owner regions for the tile cache, since we share these
+ * 	  between threads
  */
 
 /*
@@ -693,6 +696,7 @@ find_new_tile( PyramidLayer *layer )
 			if( !(layer->tiles[i].tile = 
 				im_region_create( layer->tw->im )) )
 				return( -1 );
+			im__region_no_ownership( layer->tiles[i].tile );
 			return( i );
 		}