[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[netCDF #ZCH-634698]: HDF5 error

Hi Huaiyu,

> This bug seems to be related to an earlier discussion I had with Russ
> and Ed about ticket #RUF-772159. I enabled logging in ncdump
> (4.1-beta2) and get the following error message:
> ===========================================
> ...
> read_var: obj_name position
> read_var: about to iterate over scales for dim 0
> read_var: about to iterate over scales for dim 1
> ERROR: file nc4file.c, line 1275.
> HDF error
> ERROR: file nc4file.c, line 1449.
> HDF error
> ERROR: file nc4file.c, line 1701.
> HDF error
> ===========================================
> Here variable "position" has 4 dimensions, "snapshot" (unlimited),
> "slice", "particle" and "axis". My code also defines variables
> "snapshot", "slice", "particle" and "axis" with the corresponding
> dimension. It turns out that the netcdf file becomes unreadable if
> these dimension variables are defined after "position". There are
> several strange things about this bug:
> (1) Netcdf does not report any error when the file was created.
> Actually data can be written while the program is running.
> (2) Although the file cannot be read by a ncdump which reports HDF
> error, it can be read by HDF utility program such as hdfview.
> (3) The error goes away if the "slice" dimension is removed. In
> addition, I cannot reproduce the error in a simple example code.
> 
> I know it is very difficult to catch this bug. Again I can send you
> the netcdf file if you need it.

Yes, please make the netCDF file available to us.  Perhaps with h5dump or
hdfview, we can figure out what the problem is.  Thanks.

--Russ

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



Ticket Details
===================
Ticket ID: ZCH-634698
Department: Support netCDF
Priority: Normal
Status: Closed