Re: 950321: Solaris, nctest and C++, bleah

>Keywords: 199503211819.AA28912

Hi,

> I'm having problems with netCDF under Solaris.  My config.status is included 
> at the end of this message.

I've checked your config.status file and it looks fine.  It matches exactly
what is generated here on our Solaris 2.3 platform except for the pathnames
derived from the --prefix argument you gave to configure.

> I'm having problems getting netcdf things to work under the Solaris platform.
> I have applied both patches 3 and 4 to my netcdf-232pl2 distribution,
> running configure and remaking the distribution between each.
> 
>     88  12:40   patch < 2.3.2-patch3
>     91  12:42   configure --prefix=/home/dachshund/users/fraley/netcdf\
>                               /netcdf-232pl2/dist
>     92  12:43   make all > & ! log &
>     93  12:43   tail -f log
>     96  12:49   patch < 2.3.2-patch4
>     97  12:50   configure --prefix=/home/dachshund/users/fraley/netcdf/\
>                       netcdf-232pl2/dist
>     98  12:50   make all > & ! log2 &
>     99  12:50   tail -f log2

There is no need to run `make' between applying patches 3 and 4.  You should
be able to apply patch4 immediately after applying patch3 and then run the
`configure' and `make all' steps.  You should also verify that you didn't
get any errors when invoking `patch'.  You haven't said what directory you
are running `patch' from, but to be consistent with the --prefix you are
giving `configure', I assume all the commands above, including the `patch'
commands, are being invoked from within the directory you just patched:

    /home/dachshund/users/fraley/netcdf/netcdf-232pl2/dist

This is important, because below you indicate a different problem (compiling
the C++ interface) that can be explained by your changing to the wrong
directory before you run a `make' command.

> >>>>>>>>>>>>>>>>>>>>>>
> Here is my environment:
> >>>>>>>>>>>>>>>>>>>>>>
> 
>       acs:ridgeback-> uname -a
>       SunOS ridgeback 5.3 Generic_101318-41 sun4m sparc
> 
>       acs:ridgeback-> cc -V
>       cc: SC3.0 15 Dec 1993
> 
>       acs:ridgeback-> CC -V junk.cc
>       CC: SC3.0 15 Dec 1993
>       ccfe: SC3.0 15 Dec 1993 C++4.0

Ours is the same except for the C and C++ compiler versions:

        cc: SC3.0.1 13 Jul 1994

        CC: SC3.0.1 13 Jul 1994

        ccfe: SC3.0.1 13 Jul 1994 C++4.0.1

> >>>>>>>>>>>>>>>>>>>>>>
> Here is the testing error I get...
> >>>>>>>>>>>>>>>>>>>>>>
> 
>       *** Testing ncdimdef ...        ok ***
>       *** Testing ncdimid ...         ok ***
>       *** Testing ncdiminq ...        ok ***
>       *** Testing ncdimrename ...     ok ***
>       *** Testing ncvardef ...        *** test_ncvardef: unwritten short 
>                                                       not FILL_SHORT
>       *** test_ncvardef: unwritten long not FILL_LONG
>       *** test_ncvardef: unwritten float not FILL_FLOAT
>       *** test_ncvardef: unwritten double not FILL_DOUBLE
>       *** test_ncvardef: unwritten double not FILL_DOUBLE
>       FAILED! ***
>       *** Testing ncvarid ...         ok ***
>       *** Testing ncvarinq ...        ok ***
>       *** Testing ncvarput1 ...       ok ***
>       *** Testing ncvarget1 ...       ok ***

I assume this is an excerpt from the output of running `make test' after the
last `make all' command above finished, and from the same
(.../netcdf-232pl2/dist/) directory.  I also assume no compiler or link
errors were encountered from the `make all' except multiple occurrences of
the ignorable warnings:

        semantics of ">" change in ANSI C; use explicit cast
and
        semantics of ">=" change in ANSI C; use explicit cast

I also assume the `test_ncvardef: unwritten short not FILL_SHORT' you saw is
the *first* error message produced by `make test', and in particular you
don't see any earlier errors with the XDR test, the netCDF formatted test,
or the netCDF binary test.

If any of these assumptions are wrong, please let us know, because otherwise
our conclusions might be incorrect.

We don't see the errors you saw when running `make test', but get
instead:

    making `test' in directory /a/zero/home/russ/build/netcdf-232pl2/xdr

    cc -c -I/usr/include/rpc -DNDEBUG -O xdrtest.c
    cc -O xdrtest.o xdr.o xdrfloat.o xdrstdio.o xdrarray.o -o xdrtest
    ./xdrtest > xdrtest_out.new
    diff xdrtest_out.new testout.sav
    *** XDR passes formatted test ***
    cmp test.xdr test_xdr.sav
    *** XDR passes binary test ***

    returning to directory /a/zero/home/russ/build/netcdf-232pl2


    making `test' in directory /a/zero/home/russ/build/netcdf-232pl2/libsrc

    cc -c   -DNDEBUG -O cdftest.c
    ar rcuv libnetcdf.a array.o attr.o cdf.o dim.o file.o iarray.o error.o  
globdef.o putget.o putgetg.o sharray.o string.o var.o  xdrposix.o
    cc -O cdftest.o libnetcdf.a -o cdftest
    ./cdftest > cdftest_out.new
    diff cdftest_out.new testout.sav
    *** netCDF passes formatted test ***
    cmp test.cdf test_cdf.sav
    *** netCDF passes binary test ***

    returning to directory /a/zero/home/russ/build/netcdf-232pl2


    making `test' in directory /a/zero/home/russ/build/netcdf-232pl2/nctest

    ./nctest
    *** Testing nccreate ...    ok ***
    *** Testing ncopen ...      ok ***
    *** Testing ncredef ...     ok ***
    *** Testing ncendef ...     ok ***
    *** Testing ncclose ...     ok ***
    *** Testing ncinquire ...   ok ***
    *** Testing ncsync ...      ok ***
    *** Testing ncabort ...     ok ***
    *** Testing ncdimdef ...    ok ***
    *** Testing ncdimid ...     ok ***
    *** Testing ncdiminq ...    ok ***
    *** Testing ncdimrename ... ok ***
    *** Testing ncvardef ...    ok ***
    *** Testing ncvarid ...     ok ***
    *** Testing ncvarinq ...    ok ***
    *** Testing ncvarput1 ...   ok ***
    *** Testing ncvarget1 ...   ok ***
    *** Testing ncvarput ...    ok ***
    *** Testing ncvarget ...    ok ***
    *** Testing ncvarputg ...   ok ***
    *** Testing ncvargetg ...   ok ***
    *** Testing ncrecinq ...    ok ***
    *** Testing ncrecput ...    ok ***
    *** Testing ncrecget ...    ok ***
    *** Testing ncvarrename ... ok ***
    *** Testing ncattput ...    ok ***
    *** Testing ncattinq ...    ok ***
    *** Testing ncattget ...    ok ***
    *** Testing ncattcopy ...   ok ***
    *** Testing ncattname ...   ok ***
    *** Testing ncattrename ... ok ***
    *** Testing ncattdel ...    ok ***
    *** Testing nctypelen ...   ok ***

Unless it's a bug we don't remember seeing in the earlier version of the
Solaris compiler/libraries you are using, we can't explain the discrepancy.
We no longer have any access to the 3.0 version of the C compiler, so we
can't test whether that is the problem.

If you could run `make distclean' to start again with a clean (patched)
distribution, then send us the *complete* output from running `configure',
and `make all test' from the same directory, we might be able to see
something else that would explain this puzzling problem.  No one else has
reported this problem on Solaris 2.3, and several users have successfully
built and are using netCDF on that platform.

> >>>>>>>>>>>>>>>>>>>>>>
> I can't even *compile* the C++ stuff, I'd like to use "CC"
> >>>>>>>>>>>>>>>>>>>>>>
> 
> acs:ridgeback-> pwd
> /home/dachshund/users/fraley/netcdf/netcdf-232pl2/c++

Above, you indicate that the patched version is in 

    /home/dachshund/users/fraley/netcdf/netcdf-232pl2/dist

If that's the case, it looks like you are in the wrong directory to be
running this test from.  The patches are *required* for the C++ interface.
If you run this from the patched version, "CC" should work fine, so it looks
like you should have run this from

    /home/dachshund/users/fraley/netcdf/netcdf-232pl2/dist/c++

instead, and then you wouldn't have seen these compiler errors:

> acs:ridgeback-> make
> CC -g -I../dist/include   -c  netcdf.cc
> "ncvalues.hh", line 20: Error: Multiple declaration for ncbyte.
> ...

Also, the "-I../dist/include" that appears above could only have come from a
manually modified Makefile, since the c++/Makefile in netcdf 2.3.2 uses
`-I../libsrc' instead.  The former would only work after a `make install'
had been done from the *patched* distribution.

______________________________________________________________________________

Russ Rew                                                UCAR Unidata Program
address@hidden                                          P.O. Box 3000
http://www.unidata.ucar.edu/                          Boulder, CO 80307-3000
______________________________________________________________________________