oops

TODO

@@ -1,72 +1,3 @@
-- getpoint() doesn't use a worker, it just calls vips_region_prepare()
-  directly
-
-  we should crop to a mem buffer instead
-
-- rework buffer.c, it's getting ugly
-
-  how about getting rid of buffercache for non-worker threads, it won't be
-  very useful
-
-  instead, global buffers have no buffercache and their lifetime is just
-  managed by the regions they are on
-
-  no reserve list, just discard them on unref
-
-  shouldwe use atomic stuff for the buffer ref count?
-
-  buffer_dump has:
-
-	if( buffer->im &&
-		buffer->buf &&
-		buffer->cache ) {  
-		printf( "buffer %p, %.3g MB\n", 
-			buffer, buffer->bsize / (1024 * 1024.0) ); 
-		*alive += buffer->bsize;
-	}
-	else if( buffer->im &&
-		buffer->buf &&
-		!buffer->cache )   	
-		*reserve += buffer->bsize;
-	else
-		printf( "buffer craziness!\n" ); 
-
-  is this right? 
-  
-  I think we now have three states:
-
-  global buffer (not on any cache, undone)
-
-	buffer->ref_count >= 1 
-  	buffer->im
-	buffer->buf
-	!buffer->cache
-	!buffer->done (since all done buffers must be on a cache buffers list)
-
-  buffer published on a cache (done)
-
-	buffer->ref_count >= 1 
-  	buffer->im
-	buffer->buf
-	buffer->cache
-	buffer->done 
-
-  buffer held in reserve on a thread, ready for reuse
-
-	buffer->ref_count == 0 
-  	buffer->im
-	buffer->buf
-	buffer->cache
-	!buffer->done 
-
-  vips_buffer_done() on a global buffer takes ownership and publishes to this 
-  thread's public list
-
-     what if ref_count > 1? is this possible?
-
-  vips_buffer_undone() relinquishes ownership and makes it global
-
-
 - not sure about utf8 error messages on win
 
 - strange: