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

[netCDF #THR-976469]: netcdf4: make check - Unexpected result,

> Hello Ed,
>
> thanks a lot for your immediate answer!
>
> Here the results of "make check" in the directories:
>
> ncgen:    All 2 tests passed
> ncdump:   All 22 tests passed
> nc_test:  1 of 5 tests failed (see attachment for more details)
> ### 6 FAILURES TESTING nc_put_var_float! ###
> ### 4 FAILURES TESTING nc_put_var1_float! ###
> ### 9 FAILURES TESTING nc_put_vara_float! ###
> ### 10 FAILURES TESTING nc_put_vars_float! ###
> ### 6 FAILURES TESTING nc_put_varm_float! ###
> ### 2 FAILURES TESTING nc_put_att_float! ###
> FAILURE at line(s) 712, 1329, 2855, 2902, 4077, 4080, 5335,
> 5338, 6658 of test_put.c: ...
> ->  see attachment for more details
> nctest:   All 2 tests passed
> nf_test:  All 16 tests passed
> nc_test4: All 1 tests passed
> cxx:      All 4 tests behaved as expected (1 expected failures)
> examples: All 7 tests passed
>
> I also included some print statments in the c program "tst_atts.c"
> called "make check" in the subdirectory ".../libsrc4" with the following
> result:
>
> *** Testing netcdf-4 attribute functions.
> *** testing really simple global atts...ok.
> *** testing simple global atts...ok.
> *** testing attribute data type conversions...
> **Fe: ATT_LEN = 3
> **Fe: print values in loop; loop index 0 :
> **Fe: print values in loop; loop index 1 :
> **Fe: print values in loop; loop index 2 :
> **Fe: float_in[i]: 845300.687500
> (float)double_out[i]: 0.020000
> 1 failures
> *** testing simple variable atts...ok.
> *** testing zero-length attributes... this test skipped for HDF 1.8.0
> beta1ok.
> *** testing zero-length attributes and redef...(this test skipped for
> HDF5-1.8.0 beta1ok.
> *** testing attribute deletes and renames...ok.
> *** testing attribute create order...ok.
> *** testing attribute ordering some more...ok.
> *** testing attribute ordering even more...ok.
> *** testing attributes and enddef/redef...ok.
> 1 errors detected! Sorry!
> FAIL: tst_atts
>
>
> Actually, the value of "(float)double_out[i]" always seems to be
> "0.020000", each time I run the test. The value of "float_in[i]" is
> different each time.
> It is strange for me that no values are listed for loop index 0 and 1.
>
>

Something very funny is going on here.

double_out is set a little bit above this part of the code:
double_out[ATT_LEN] = {0.02, .1128, 1090.1};

So what is up?

I would like you to try a very simple build, and see if we encounter this
problem. Since you build HDF5 with icc, why are you building netCDF with mpicc?
If you want parallel I/O you must build HDF5 with mpicc also.

Try building netCDF like this and, if it doesn't work, send me the complete
output of configure and make, as well as the config.log file. Just go ahead and
paste the output in the body of your email, it's OK if it's very long.

Do this:

make distclean
CC=icc ./configure --enable-netcdf-4 --with-hdf5=/wherever --disable-f77
--disable-cxx
make check

(Of course set whatever other environmental variables icc needs. These
configure options turn off all fortran APIs and tests, so you don't have to
bother about fortran for this build.)

Please let me know if this works.



> By the way: why does make check tests "64-bit offset format" even thouch
> it is installed on a 32-bit Linux system?
>
> Thanks,
> Notker
>

64-bit offset format is one of the underlying binary formats of netCDF (and
works whether you have a 32-bit or 64-bit processor). By using 64-bit long
internal offsets it allows very much larger data sizes.

For more on netcdf binary formats see:
http://www.unidata.ucar.edu/software/netcdf/docs/netcdf/File-Format.html#File-Format

Thanks,

Ed

Ticket Details
===================
Ticket ID: THR-976469
Department: Support netCDF
Priority: Normal
Status: Open