nc_test4 failed when using intel compilers

[dongli]

Please provide as much of the following information as you can, as applicable to the issue being reported. Naturally, not all information is relevant to every issue, but the more information we have to start, the better!

Environment Information

Feel free to skip this if the issue is related to documentation, a feature request, or general discussion.

• What platform are you using? (please provide specific distribution/version in summary)
  • Linux
  • Windows
  • OSX
  • Other
  • NA
• 32 and/or 64 bit?
  • 32-bit
  • 64-bit
• What build system are you using?
  • autotools (configure)
  • cmake
• Can you provide a sample netCDF file or C code to recreate the issue?
  • Yes (please attach to this issue, thank you!)
  • No
  • Not at this time

Summary of Issue

I am install netcdf-c in a CentOS 7.4.1708 system with the following compilers:

• C: icc version 13.1.0 (gcc version 4.7.0 compatibility)
• Fortran: ifort version 13.1.0

The dependent zlib, szip, hdf5 (1.10.1) have already be installed. The make is done, but make check gave the following errors from nc_test4/test-suite.log:

===========================================
   netCDF 4.6.0: nc_test4/test-suite.log
===========================================

# TOTAL: 66
# PASS:  64
# SKIP:  0
# XFAIL: 0
# FAIL:  2
# XPASS: 0
# ERROR: 0

.. contents:: :depth: 2

FAIL: tst_filterparser
======================

mismatch: N.A. [7] baseline=0d params=3287505826
mismatch: N.A. [8] baseline=0d params=1097305129
mismatch2: ud.d [7-8] baseline=0d,0d params=3287505826,1097305129

Testing filter parser.
FAIL tst_filterparser (exit status: 1)

FAIL: tst_filter
================

findplugin.sh loaded
final HDF5_PLUGIN_PATH=/tmp/starman/netcdf-c/netcdf-4.6.0/nc_test4/hdf5plugins/.libs
*** Testing dynamic filters using API

*** Testing API: bzip2 compression.
show parameters for bzip2: level=9
show chunks: chunks=4,4,4,4

*** Testing API: bzip2 decompression.
data comparison: |array|=256
no data errors
*** Pass: API dynamic filter

*** Testing dynamic filters parameter passing
test1: compression.
test: nparams=14: params= 1 239 23 65511 27 77 93 1145389056 0 0 1 2147483648 4294967295 4294967295
 chunks=4,4,4,4
test1: decompression.
data comparison: |array|=256
no data errors
1c1
< var:_Filter = "32768,1,239,23,65511,27,77,93,1145389056,0,0,1,2147483648,4294967295,4294967295" ;
---
> var:_Filter = "32768,1,239,23,65511,27,77,93,1145389056,3287505826,1097305129,1,2147483648,4294967295,4294967295" ;
FAIL tst_filter.sh (exit status: 1)

Steps to reproduce the behavior

[DennisHeimbigner]

Are you by any chance running this on a big-endian machine?

[dongli]

@DennisHeimbigner Sorry, I have no access to big-endian machines.

[DennisHeimbigner]

Good, then we have a chance of solving the problem :-)

[justbennet]

I am reporting the same issue. Using Intel compilers 14.0.2, hdf5 1.8.20, CentOS 7.5. The configure line used was

./configure --prefix=/sw/arcts/centos7/intel_14_0_2/netcdf/4.6.0

All tests pass except the two filter tests.

[bennet@beta-build nc_test4]$ ./tst_filter.sh
findplugin.sh loaded
final HDF5_PLUGIN_PATH=/tmp/build/netcdf-4.6.0/nc_test4/hdf5plugins/.libs
*** Testing dynamic filters using API

*** Testing API: bzip2 compression.
show parameters for bzip2: level=9
show chunks: chunks=4,4,4,4

*** Testing API: bzip2 decompression.
data comparison: |array|=256
no data errors
*** Pass: API dynamic filter

*** Testing dynamic filters parameter passing
test1: compression.
test: nparams=14: params= 1 239 23 65511 27 77 93 1145389056 0 0 1 2147483648 4294967295 4294967295
 chunks=4,4,4,4
test1: decompression.
data comparison: |array|=256
no data errors
1c1
< var:_Filter = "32768,1,239,23,65511,27,77,93,1145389056,0,0,1,2147483648,4294967295,4294967295" ;
---
> var:_Filter = "32768,1,239,23,65511,27,77,93,1145389056,3287505826,1097305129,1,2147483648,4294967295,4294967295" ;

Output from the executables directly.

[bennet@beta-build nc_test4]$ ./test_filter

*** Testing API: bzip2 compression.
show parameters for bzip2: level=9
show chunks: chunks=4,4,4,4
fail (218): NetCDF: HDF error

[bennet@beta-build nc_test4]$ ./tst_filterparser 

Testing filter parser.
mismatch: N.A. [7] baseline=0d params=3287505826
mismatch: N.A. [8] baseline=0d params=1097305129
mismatch2: ud.d [7-8] baseline=0d,0d params=3287505826,1097305129

[DennisHeimbigner]

I looked again at the Intel web-site and it appears that there is still no way for Unidata
to obtain a free copy of the Intel icc compiler.
So I doubt that we will be able to solve this.
My only suggestion is to try reducing the level of optimization you using and
see if that fixes the problem.

[justbennet]

Thanks for the suggestion about optimization. I will try with -O0,

I also looked more closely at the log, it may be an error in the underlying hdf5, which I will try to trace.

[justbennet]

I looked at test_filter.c, and it appears to fail at

    CHECK(nc_enddef(ncid));

Is this by any chance using a feature of HDF5 that was introduced after 1.8.20? It appears to be getting an error from HDF5.

[DennisHeimbigner]

It is more likely that the netcdf-c library is calling the HDF5 API with some bad value.
The fact that it is at enddef indicates that it is a metadata error as opposed to
a data error.
We may be able to get more info if you do the following.

1. rebuild the netcdf library with the extra ./configure options --enable-logging
2. set this environment variable
   export NETCDF_LOG_LEVEL=5
   3.rerun the test (with the -x flag to sh).

[justbennet]

Dennis, I will try that. This will take me a day or so to complete, but I will update you.

…

-- bennet

On Sun, Oct 7, 2018 at 2:31 PM Dennis Heimbigner ***@***.***> wrote: It is more likely that the netcdf-c library is calling the HDF5 API with some bad value. The fact that it is at enddef indicates that it is a metadata error as opposed to a data error. We may be able to get more info if you do the following. 1. rebuild the netcdf library with the extra ./configure options --enable-logging 2. set this environment variable export NETCDF_LOG_LEVEL=5 3.rerun the test (with the -x flag to sh). — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#876 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABhYwoP8K0ePF5Ls_iWhAo1fMK7Keb6gks5uikhygaJpZM4SNueE> .

[liubiyongge]

I meet the same error, could you tell me how to solve this problem?