[svn-r323] Changes since 19980316

----------------------

./src/H5D.c
./src/H5T.c
./src/H5Tprivate.h
	Fixed a bug found by Kevin Powell regarding preservation of
	existing data during a read or write when the source and
	destination data types are the same.  Thanks Kevin!

	Fixed a couple warnings on Irix64.

./src/H5D.c
./src/H5Dprivate.h
./src/H5P.c
./src/H5Ppublic.c
./src/H5Tconv.c
./src/H5Tpublic.h
./test/cmpd_dset.c
	The application can now turn on/off the part of the I/O
	pipeline that deals with preservation of initialized data.
	The default is off since this will be the usual case and
	turning it on slows down the pipe.  Use H5Pset_preserve() and
	H5Pget_preserve().

./src/H5Fistore.c
	Added an optimization for reading/writing a single chunk of
	chunked storage.

./src/H5Odtype.c
./src/H5Oefl.c
./src/H5Olayout.c
./src/H5Oname.c
./src/H5Osdspace.c
./src/H5Ostab.c
	Fixed warnings reported by marek@iiasa.ac.at.  Thanks Marek.

./src/H5Odtype.c
	Fixed a failing assert() wrt reading compound types (not
	present in hdf5-1.0.0a).

1 files changed, 1 insertions, 9 deletions

diff --git a/src/H5Tconv.c b/src/H5Tconv.c
index 604569c..beae526 100644
--- a/src/H5Tconv.c
+++ b/src/H5Tconv.c
@@ -297,15 +297,7 @@ H5T_conv_struct_init (H5T_t *src, H5T_t *dst, H5T_cdata_t *cdata)
 	}
     }
 
-#ifndef LATER
-    /*
-     * Always use an initialized background buffer.  Actually, we usually
-     * won't need it initialized but we'll almost always need a buffer to
-     * move the members around.
-     */
-    cdata->need_bkg = H5T_BKG_YES;
-#endif
-
+    cdata->need_bkg = H5T_BKG_TEMP;
     cdata->recalc = FALSE;
     FUNC_LEAVE (SUCCEED);
 }