Hi Mary,
> Is the recent HDF5 1.8.4 patch all that's needed to get NetCDF 4.1
> working again, or is a patch required for NetCDF 4.1 as well?
The bug writing netCDF-4 files on big-endian platforms with more than 8
attributes attached to a variable or group is described here:
http://www.hdfgroup.org/HDF5/release/known_problems/index.html
That bug is fixed by release HDF5 1.8.4 Patch 1, available from the HDF5
downloads page:
http://www.hdfgroup.org/downloads/index.html
Building *any* version of netCDF-4 with HDF5 1.8.4 Patch 1 fixes that
bug.
The unrelated bug in the netCDF 4.1 release, reported on this mailing
list by Jeff Whitaker on Feb 13, requires the conditions described by Ed
here:
http://www.unidata.ucar.edu/mailing_lists/archives/netcdfgroup/2010/msg00070.html
That bug is fixed in the netCDF-4.1.1 release. We will soon be
announcing the availability of release candidate netcdf-4.1.1-rc1, which
will become the official 4.1.1 release a week or so later if no new
significant bugs are reported.
The netCDF-4.1.1 release also contains a test for the HDF5 bug fixed in
1.8.4 patch 1, and a problem with that test is all that's currently
holding things up.
> I'm wondering if I should redo all my NCL builds now, or wait for a
> NetCDF 4.1. patch.
I recommend building with HDF5-1.8.4 patch 1 to anyone currently writing
netCDF-4 files on big-endian platforms. Otherwise, I recommend waiting
for the netCDF-4.1.1-rc1 or 4.1.1 release and building that with the
HDF5-1.8.4 Patch 1 release.
--Russ
