Hi all,
If any of you are using a beta version of 5.2.0 that we provided you
with, please be aware that the Unidata/HDF5 folks have uncovered a
serious bug that could manifest itself if you use this NCL to create
NetCDF files:
We have just been informed of a bug in HDF5-1.8.4 that causes a
problem
in files written using netCDF-4.1. The symptom is that if a
variable or
group is created with more than 8 attributes, the resulting file
becomes
non-portable between little-endian and big-endian platforms.
The HDF5 developers have a fix and are testing it now. We are also
about to test a prerelease snapshot of the fix.
This note is to inform the netCDF community about this problem and
the
upcoming fix, which we expect to be officially announced sometime
next
week.
In the meantime, we recommend that you not use netCDF-4 linked with
HDF5-1.8.4 to write netCDF-4 or netCDF-4 classic model files that
have
more than 8 attributes for any data object, including more than 8
global
attributes. We will add a test to the netCDF-4 version 4.1.1 release
that would have uncovered this bug, test the HDF5 fix to make sure it
creates portable netCDF-4 files, and keep you informed of progress on
this issue.
I built the NCL 5.2.0 beta binaries against these versions of NetCDF
and HDF5, so I recommend that you don't use NCL 5.2.0 beta to write
any NetCDF files.
I'll let you know when/if new 5.2.0 beta binaries become available.
--Mary
_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Mon Feb 22 13:34:30 2010
This archive was generated by hypermail 2.1.8 : Tue Feb 23 2010 - 08:26:41 MST