Due to the current gap in continued funding from the U.S. National Science Foundation (NSF), the NSF Unidata Program Center has temporarily paused most operations. See NSF Unidata Pause in Most Operations for details.
Due to the current gap in continued funding from the U.S. National Science Foundation (NSF), the NSF Unidata Program Center has temporarily paused most operations. See NSF Unidata Pause in Most Operations for details.
This is IBM poe under AIX. On Mon, Jan 18, 2010 at 3:29 PM, Ted Mansell <Ted.Mansell@xxxxxxxx> wrote: > Which flavor of MPI are you using? There is a problem with OpenMPI that, > as far as I know, has not been fixed in the netcdf4 release. The error > sounds similar. If that is the problem, there was a fix sent to the list > sometime last fall, I think. (It worked for me, at least.) > > -- Ted > > > On Jan 18, 2010, at 3:51 PM, Jim Edwards wrote: > > I am recently porting my code to netcdf4 and think that there may be a bug >> in the underlying hdf5 library... >> >> I'm not sure where to take it, thought I would try this list first. >> >> The issue is that when a file is opened for read using nf90_open_par I get >> an error on closing the file, I managed to get a traceback by using the mpi >> wrappers facility >> >> 1:call chain trace-back: >> 1: MPI_Comm_free >> 1: H5FD_mpi_comm_info_free >> 1: H5FD_mpio_close >> 1: H5FD_close >> 1: H5F_dest >> 1: H5F_try_close >> 1: H5F_close >> 1: H5I_dec_ref >> 1: H5Fclose >> 1: close_netcdf4_file >> 1: nc_close >> 1: nf_close >> 1: __netcdf_NMOD_nf90_close >> >> >> 1:ERROR: 0032-131 Can't free a predefined communicator (0) in >> MPI_Comm_free, >> >> This is using netcdf 4.0.1 and hdf5 1.8.3 >> >> I don't seem to be getting this error on a file opened to write, only >> those opened to read. >> >
netcdfgroup archives: