20041122: possible bug in NetCDF 3.6 beta6 on Itanium Linux

Howard,

> To: address@hidden
> From: "Howard M. Salis" <address@hidden>
> Subject: Bug in NetCDF 3.6 beta6 netcdf_mp_nf90_get_var_1d_fourbyteint on 
> Itanium Linux
> Organization: UCAR/Unidata
> Keywords: 200411201943.iAKJhN7j020088

The above message contained the following:

>   I compiled NetCDF 3.6 beta 6 on an Itanium (64-bit) Linux machine
> successfully (make test succeeded), but when I attempted to read a NetCDF
> file one particular NetCDF function
> 'netcdf_mp_nf90_get_var_1d_fourbyteint' died in a segmentation fault.
> 
>   I've doublechecked to make sure the NetCDF variable being read is an
> array of small integer numbers.
> 
>   I'm using the Intel C/C++ and Fortran compilers and I've included the
> -mp flags for both (as suggested in an archived support email solving a
> previous problem someone else experienced). Because the cfortran.h file
> doesn't recognize the Intel compilers, I've had to overwrite the existing
> cfortran.h with the patched version of NetCDF 3.5's. No one has released a
> patch for 3.6's cfortran.h (the line numbers were changed so the old patch
> fails).
> 
>   I've installed NetCDF 3.5 on a different Itanium64 machine and I had no
> problems.

Did you try to duplicate the problem on the other Itanium64 machine
using version 3.5 of the netCDF package?  If so, what was the result?

>   I'm including the configure.log, config.log, make.log, test.log, and
> install.log files as you request.

Those files indicate that the netCDF package was successfully built and
passed all its tests.

> I've also included the output of uname
> -a, the locations of the compilers, and the output of Intel's debugger,
> idb, when I run my program. The program is called 'NHybrid-Debug' and the
> bug occurs in file 'DataInput-new.f' in function 'InputModelData' in
> module 'DataInput'.

Can you supply us with a simple program that demonstrates the problem so
that we can try to reproduce it?

If you can't, then you might have to debug the problem yourself.  I
would start by rebuilding the netCDF package and your program using the
"-g" (debug) option.

Alternatively, can you give us access to your system?

> I'd appreciate any help you could provide. I'm trying to get NetCDF
> working on one of NCSA's Teragrid machines. I guess I can always go back
> to v3.5, but I'd like to try out newer versions of NetCDF. Thank you.
> 
> Sincerely,
>    Howard Salis
>    Dept Chemical Engineering & Materials Science
>    University of Minnesota
> 
> 
> Linux tg-login1 2.4.21.SuSE_241.bef1 #1 SMP Thu Aug 12 12:03:48 CDT 2004
> ia64 unknown
> 
> /opt/intel/compiler80/bin/ifort
> /opt/intel/compiler80/bin/icc
> 
> Thread received signal SEGV stopped at [<opaque> function
> netcdf_mp_nf90_get_var_1d_fourbyteint_(...)  0x400000000008e350]
> Information:  An <opaque> type was presented during execution of the
> previous command.  For complete type information on this symbol,
> recompilation of the program will be necessary.  Consult the compiler man
> pages for details on producing full symbol table information using the
> '-g' (and '-gall' for cxx) flags. (idb)
> 
> where
> 
> >0 0x400000000008e350 in netcdf_mp_nf90_get_var_1d_fourbyteint_(...) in
> NHybrid-Debug
> 
> #1 0x4000000000132d20 in DATAINPUT::inputmodeldata() "DataInput-new.f":268
> 
> #2 0x4000000000191930 in mainprogram()
> "mainprogramH-new.f":28
> 
> #3 0x4000000000004210 in main(...) in
> NHybrid-Debug
> 
> #4 0x20000000001f3fa0 in __libc_start_main(...) in
> /lib/libc.so.6.1
> 
> #5  0x4000000000003c40 in _start(...) in NHybrid-Debug
...

Regards,
Steve Emmerson

> 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.