Re: 970207: nctest failes, suggestions?

>To: address@hidden
>From: "Kent F. Milfeld" <address@hidden>
>Subject: nctest failes, suggestions?
>Organization: .
>Keywords: 199702071300.GAA25689

Hi Kent,

First, apologies for not replying to this question earlier.  We were all
out of town at a meeting last week and came back to a large backlog of
support questions ...

>    On our Cray J90 the test section of the Make (make test) keeps failing
> when I try different compilers (CC=cc, CC=g89).  It appears that nccreate
> makes a file, with the contents "CDF^A" , and then bombs when it tries to
> open it.  Here is the error information:
> 
> c89 -o nctest -O varget.o vargetg.o varput.o varputg.o vardef.o \
>                  vartests.o vputget.o vputgetg.o driver.o cdftests.o \
>                  dimtests.o rec.o atttests.o misctest.o add.o error.o \
>                  emalloc.o val.o slabs.o ../libsrc/libnetcdf.a \
>       ./nctest
> Error Explanation:Make: "./nctest" terminated due to signal 11 (core dumped)
> cmd-2436 make: Stop.

I'm forwarding your message to Steve Emmerson, who has successfully
built and tested netCDF 2.4.3 on NCAR's J90.  He'll be looking at it in
the next day or two, to see if we can duplicate the problem here or see
what's wrong.

> PS  About two weeks ago I send in a problem report for the CC++ interface,
>     but did not get a reply.  On our Cray J90 the CC++ make choked on
>     two constructors which used, (I believe) char and unsigned char.  I
>     think I found a fix with the "-h char" option on the Cray CC compiler.
>     I'll try it after I get finished installing the C and Fortran interfaces
>     (I turned the CC compilations off.)

I've appended a copy of the reply I sent on Tuesday, 21 January, to the
address

   "Kent F. Milfeld" <address@hidden>

I'm sorry if you didn't get it, but we never received a "bounce" message
or any other indication that the mail failed.

Please let me know if you don't get this one :-).

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu

 To: "Kent F. Milfeld" <address@hidden>
 Cc: support-netcdf
 Subject: Re: 970121: Cray J90 CC compiler error 
 Date: Tue, 21 Jan 1997 17:02:08 -0700
 From: Russ Rew <address@hidden>

 >To: address@hidden
 >From: "Kent F. Milfeld" <address@hidden>
 >Subject: Cray J90 CC compiler error
 >Organization: .
 >Keywords: 199701212143.OAA18759

 Hi Kent,

 >    A user asked that we install netcdf on our system, a Cray J90 running
 > UNICOS 9.0.2.                  (netcdf version 2.4.3)
 > 
 >    I used the defaults for configuring the makefiles.  The C and FORTRAN
 > libraries, as well as ncgen and nctest compiled fine.  However, the 
 > C++ code gave quite a few errors, which look like they are due to one single
 >  type of occurrence.  The errors occur in  netcdf.hh.
 >   (Please forgive my ignorance of C++; I am a FORTRAN "person".)
 > 
 > The errors appear to occur when "char" and "ncbyte" are used in successive
 > constructors.  The compiler believes that the constructors are identical,
 > and reports an error when ncbyte is used after char, or vice versa. 
 > For example, the first error occurs for these two lines:
 > 
 >     NcBool add_att( NcToken attname, char );             // scalar attributes
 >     NcBool add_att( NcToken attname, ncbyte );
 > 
 > with the message:
 > CC-403 CC: ERROR File = netcdf.hh, Line = 76
 >   Member function function "NcFile::add_att" has already been declared.
 > 
 > Other errors related to the use of char and ncbyte occur throughout the code.
 > Below is a copy of the makefile and the make.log.  If you have a workaround
 > or fix, could you please advise.  

 We discuss a fix in our document "Known Problems with the netCDF
 Distribution", although this problem is really a Cray C++ compiler bug.
 The workaround description is at

     http://www.unidata.ucar.edu/packages/netcdf/known_problems.html#cray-cxx

 and a previous message/reply that gives more details about the problem
 and workaround, in case you're interested, is available at

     http://www.unidata.ucar.edu/glimpse/netcdf/2009

 > Also, config.log showed a couple of errors in conftest.c, I have
 > included them at the end.  Any suggestions?

 Errors in config.log may be ignored, since they are merely side-effects
 of the "configure" script trying to determine the correct configuration
 by compiling various test programs and noting compiler/linker errors.

 _____________________________________________________________________

 Russ Rew                                         UCAR Unidata Program
 address@hidden                     http://www.unidata.ucar.edu