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

Re: netCDF on the Paragon

> Organization: ORNL
> Keywords: 199403142143.AA15685

Hi Tim,

> I've been trying unsuccessfully to get netCDF up and running on our
> Paragon systems at ORNL. I'm hoping you can help me.
> 
> The Paragons run a version of OSF.
>
> I have modified the CONFIGURE file as follows:
> 
> OS=osf
> FC=f77
> LD_XDR='-L/lib -lrpc'
> 
> (the last line is so that the xdr library provided with the Paragon
> will be linked in.)
> 
> Running configure produces no unexpected error or warning messages.
> 
> 'make all' produces a few warnings of the type:
> 
> PGC-W-0136-Function emalloc has non-prototype declaration in scope 
> (./emalloc.h: 18)
> 
> and
> 
> PGC-W-0095-Type cast required for this conversion (xdrposix.c: 132)
> 
> but I don't believe that either of these would cause any problems in
> the compiled code.

I haven't seen these warnings before on other platforms, but I tend to agree
that they shouldn't be the cause of the problems you are seeing.

> lex produces the following:
> 
> lex ncgen.l
> 28: (Warning) Definition nonquotes not found
> 82: (Warning) Definition exp not found
> 82: (Warning) Definition exp not found
> 89: (Warning) Definition exp not found
> 89: (Warning) Definition exp not found

I don't understand these warnings and haven't seen them on any other
platforms.  The lex definitions for escquote, nonquotes, and exp are in
lines 22-24 of ncgen.l.  This appears to be a bug in the Paragon version of
lex.  You might try moving these three definitions up to the beginning of
ncgen.l, or use a different version of lex such as flex, available in source
form from ftp://prep.ai.mit.edu/pub/gnu/flex-2.4.6.tar.gz for example.
These warnings do not cause the errors you are seeing in running the XDR
test, however; they are only relevant for building the ncgen program.

> mv lex.yy.c ncgenyy.c
> yacc -d ncgen.y
> mv y.tab.c ncgentab.c
> 
> I don't know if this could cause a problem or not.
> 
> When I do 'make test', I get the bad news:
> 
> making `test' in directory /home/xps5/u1/sheehan/netcdf/netcdf-2.3.2/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
> test.xdr test_xdr.sav differ: byte 28169, line 1
> *** XDR fails binary test ***
> *** Exit 1
> Stop.
> *** Exit 1

The comparison of binary files using "cmp" is very stringent, but
it seems to work on a wide variety of other platforms, so we haven't tried
to come up with a more forgiving test.  I'm afraid the only way to tell if
this is serious or not is to find out the source of the difference in a
debugger, but I don't have access to a Paragon platform.  The printed output
of xdrtest is apparently OK, since the formatted test was passed.  The byte
location is close to the end of the file, so I would guess that the
difference in the binary files has to do with the XDR encoding of enums,
written by the statement

        assert( xdr_vector(xdrs, (char *)encounts, count, szof, xdr_enum)) ;

in xdrtest.c.  It's possible this is telling us that the Paragon encoding of
enums in their XDR library isn't correct, or at least doesn't agree with the
way other paltforms encode enums.

> I installed netCDF through the 'make test' step, and it worked
> flawlessly. I would appreciate any insight you can give me on
> installing it on the Paragon.

A few other users have tried to install netCDF on Paragon platforms, and
some of them may have succeeded.  If you have Mosaic or gopher, try
searching for the keyword "paragon" in our email support archives in 

    http://www.unidata.ucar.edu/packages/netcdf/index.html

(a URL for a WWW clinet like Mosaic or lynx) or

    
gopher://groucho.unidata.ucar.edu/7waissrc%3a/systems/netcdf/unidata-support-netcdf.src

for gopher.  These will give you email addresses of others who have tried to
install netcdf on Paragons.

__________________________________________________________________________
                      
Russ Rew                                              UCAR Unidata Program
address@hidden                                        P.O. Box 3000
(303)497-8645                                 Boulder, Colorado 80307-3000