further nc_test progress

[Robert Latham]

On Thu, Mar 13, 2003 at 02:39:22PM -0600, Jianwei Li wrote:
> cool!
> 
> One comment while I'm looking into the nc_test code:
> 	When I was linking the nc_test, I get these errors:
> 	
> 		test_put.o: In function `test_ncmpi_put_var_short':
> 		test_put.o(.text+0x59b2): undefined reference to
> 				`ncmpi_put_var_type_all'
> 	
> 	Just for your information, ncmpi_put_var_type_all is not defined
> 	in our parallel interface since is of no use and meaningless, though
> 	'ncmpi_get_var_type_all' does exist and is intended to optimize
> 	processes all requesting a whole array (optimization not implemented
> 	yet?).

grr.. sorry about that. i didn't rebuild after a "make clean" so i
guess it linked successfully because of some stale object files (?). 

It's fixed in cvs now, but if you think it should be fixed a different
way, let me know so we can add a "how to use independent variable
calls" section to our porting guide ( even though that guide doesn't
exist quite yet :> )
 	
> I'll keep going through the tests and hope to find sth.

fantastic.  If it helps, we fail the same way even if the netcdf file
was created by the serial version of nc_test, so we can rule out
"created a bad file", even though the .nc file we create is a bit
smaller than the .nc file created by the serial version.

==rob

-- 
Rob Latham
Mathematics and Computer Science Division    A215 0178 EA2D B059 8CDF
Argonne National Labs, IL USA                B29D F333 664A 4280 315B