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

960411: CRAY J90 installation problems

Christi,

>Date: Thu, 11 Apr 1996 11:35:14 -0600 (MDT) 
>From: Christi Forsythe <address@hidden>
>Organization: Sandia Labs
>Subject: installation problems 
>Keywords: 199604111735.AA01785 

In the above message you wrote:

> Steve,
> 
> FYI: I recently sent you an email telling you that someone had told
> me that GNU was switch from autoconf to configure.  Well, I haven't
> heard or found anything to confirm this.  I downloaded a couple things
> from gnu and they use autoconf.  Anyways, rewriting our installation
> process has been put aside for a short time while cleaning up some
> existing problem have been put on the top of the priority list.
> I'll return to the installation problem later.
> 
> I've been trying to install netcdf-2.4.1 on our cray and was having some
> problems.  A while ago I talked to you about problems installing a beta
> version. I know our uname command was returning strange results.
> sacj001(55)% uname -a
> sacj001 sacj001 8.0.4 roo.6 CRAY J90
>  
> Anyways, I set the OS environment variable just like you told me to 
> do last time.
> % setenv OS unicos
> 
> and then executed the following:
> 
> % ./configure > ca.log
> 
> My CUSTOMIZE file contained the following:
> #
> # C compiler (default is `cc'):
> # CC=cc
> #
> # C compiler flags (default is `-O'):
> # CFLAGS=
> #
> # C preprocessor flags (default is empty):
> # CPPFLAGS=
> #
> # C++ compiler (default is system-dependent).  Define as the empty
> # string to avoid building the C++ interface.
> # CXX=""
> #
> # Installation prefix (default is `..'):
> # prefix=/usr/community/caforsy/lib
> 
> 
> The config.log file contains the following:
> ---
> This file contains any messages produced by compilers while
> running configure, to aid debugging if configure makes a mistake.
> 
> cc-428 c89: ERROR File = configure, Line = 2231
>   The type of "malloc" is not compatible with its previously declared type.
> 
> TOTAL ERRORS DETECTED IN conftest.c: 1
> ---

The above is Ok.  As is says, it just contains log messages from
utilities that the configure script invokes.  It's not an error for a
utility to error abort and emit a message when the configure script is
checking for something (e.g. whether or not malloc() is declared in
<stdlib.h>).

...

> My ca.log file (output from ./configure) contained the following:
> 
> loading cache ./config.cache
> checking for fill-value usage... new usage
> checking the installation prefix... /usr/community/caforsy/netcdf/netcdf-2.4.1
> checking the installation exec-prefix... 
> /usr/community/caforsy/netcdf/netcdf-2.
> 4.1
> ****QUESTION: my CUSTOMIZE file set the installation directory as
>       /usr/community/caforsy/lib  What did I do incorrectly here?

I appears that the configure script didn't read the CUSTOMIZE file at
all.  From looking at my configure script, this should only happen if the
file doesn't exist in the current directory or isn't readable.

> checking for c89... (cached) c89
> checking type of operating system... unicos
> checking type of machine... cray j90
> checking for ranlib... (cached) :
> checking how to run the C preprocessor... (cached) c89 -E
> checking the C preprocessor... works
> checking for f77... no
> checking for cf77... cf77
> checking for CC... CC
> checking for yacc... (cached) yacc
> checking for neqn... (cached) neqn
> checking for tbl... (cached) tbl
> checking for ar... (cached) ar
> checking for makeinfo... no
> checking for tar flags... -chof
> checking for dependency generation mechanism... c89 -M
> checking whether cross-compiling... (cached) no
> checking size of int... (cached) 8
> checking size of long... (cached) 8
> checking type of nclong... int
> checking endianess... big endian
> checking type of netlong... no 32-bit type: using long
> checking type of internal netlong... long
> checking for XDR header-file... -I/usr/include/rpc
> checking for XDR library...
> checking for xdr_long in default library... yes
> checking XDR implementation... ok
> checking type pointed to by inline XDR function... inline_t
> checking XDR library for xdr_getint()... no
> checking which XDR module to use... xdrffio
> checking for function prototype... yes
> checking for working const... yes
> checking for variadic function support... yes
> checking for strerror... (cached) yes
> checking compiler options for IEEE arithmetic...
> checking for size of block buffer... 196608
> checking for float.h... (cached) yes
> checking C header file <stdlib.h> for function malloc()... declared
> ****NOTE: everything looks fine here, but the config.log file had an error.
> checking for package version... 2.4
> checking binary distribution directory... /home/ftp/pub/binary/cray j90-unicos
> updating cache ./config.cache
> creating ./config.status
> creating Makefile
> creating xdr/Makefile
> creating libsrc/Makefile
> creating fortran/Makefile
> creating cxx/Makefile
> creating ncgen/Makefile
> creating ncdump/Makefile
> creating nctest/Makefile
> creating doc/Makefile
> creating port/master.mk
> creating port/Makefile
> creating libsrc/local_nc.h
> libsrc/local_nc.h is unchanged
> 
> 
> Any help would be much appreciated.

You needn't worry about the `malloc()' message in config.log.

Figure out why the configure script didn't read your CUSTOMIZE file, or,
alternatively, set the variables in the environment prior to running the
configure script.

Be sure to remove the file config.cache before re-executing the
configure script.

Otherwise, it looks fine.

Please let me know if this helps.

--------
Steve Emmerson   <address@hidden>