Hello, Thank you for providing the information below; I had read your pull request (https://github.com/Unidata/netcdf-c/pull/120) while I was out of the office last week, and was trying to figure out the context where this check would cause problems. This answers that question splendidly. I will need to check for potential unintended effects of the pull request which applies the fix you describe below, but I should be able to do that in the next day or two. Thanks again for the detailed information! -Ward > In netcdf-4.3.2 and netcdf-4.3.1, we use the capability of calling > nc_get_var_longlong to read an int dataset variable into a longlong > client datatype. This works in serial and in parallel. > > I recently tested upgrading to netcdf-4.3.3.1 and am encountering errors > when doing this. The basic issue seems to be that the dispatcher in the > 4.3.2 is NC_DISPATCH_NC3; however, in 4.3.3.1, the dispatcher is forced > to be NC_DISPATCH_NC5 when run in parallel. This is done at line 177 > of dfile.c. > > The problem with this is that when I call nc_get_var_longlong, I get an > NC_EINVAL return from line 569 in nc5dispatch.c. > > If I remove that memtype check, then the code falls through to the switch > statements and seems to correctly handle the NC_INT64 memory type at > either line 598 or line 619 in the same file. > > It looks like there is either a problem with forcing the dispatcher to be > NC_DISPATCH_NC5 or the "memtype == NC_INT64" check needs to be removed. > Note that the NC5_get_att and NC5_put_att functions do not have that > check, but the NC5_get_vara and NC5_put_vara functions do. > > Thanks, > ..Greg > > Ticket Details =================== Ticket ID: DMD-262410 Department: Support netCDF Priority: Normal Status: Closed
NOTE: All email exchanges with Unidata User Support are recorded in the Unidata inquiry tracking system and then made publicly available through the web. If you do not want to have your interactions made available in this way, you must let us know in each email you send to us.