[netCDF #RUF-772159]: group name as const char* in nc_def_grp

Huaiyu,

> BTW, do you have any idea why netCDF 4.1 can generate an invalid HDF 5 
> file sometimes? I have encountered this problem in my code, but I 
> cannot reproduce it using simpler codes. So I didn't formerly report 
> the bug. (My code checks the return status in every netCDF call, and 
> there is no complain from netCDF at all.) In the end, I get around the 
> problem by changing the order of defining variables.

The only ways I can think of that an invalid or corrupt HDF5 file could be
generated through the netCDF-4 interface are

  - not calling nc_close() on the file before exiting, which means nc_close()
     doesn't get a chance to flush HDF5 buffers and properly close the HDF5
     file

  - interrupting the netCDF-4 program before it can call nc_close() or before
     nc_close() returns successfully

  - using the parallel interfaces improperly, so that two threads or processes 
     are writing the same HDF5 file at the same time without following the
     discipline required to make sure the file isn't clobbered.

If none of these apply in your case, we would be very interested if you can
duplicate the bug consistently, so we could diagnose the cause.

--Russ

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu



Ticket Details
===================
Ticket ID: RUF-772159
Department: Support netCDF
Priority: High
Status: Closed