which I surrounded by a similar #ifndef __GNUC__
as in the two files above.
Does that make sense?
An additional thing: On Irix, the configure
script (correctly) finds getop and getopt.h
somewhere, but the compiler complains about
struct option not defined later on (e.g., in ncap.c). There is a comment in ncap.c, saying that it is unclear how SGI handles getopt; well, I'd conclude nit really well;-))
For me, undefining HAVE_GETOPT and HAVE_GETOPT_H in config.h after ./configure, but previous to make, seems to have fixed the problem.
Thanks for the tips. It's good to know the configure
mechanism works well with the GNU compilers on
these systems.
> Does that make sense?
Not to me. __GNUG__ applies only to g++, so the
tokens are only meant to be defined if g++ is not
being used. gcc does not want them on my systems.
Change to __GNUC__ should cause"symbol re-defined"
warnings or something like that. I'm afraid if I make the changes you needed to the main tree,
there will be unacceptable regressions in other
environments. We will refer people who use gcc
on SGI or HPPA to your email if the have problems
though.
Thanks for the report!
Charlie
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I just tried to install (successfully) nco-2.7.3
on both HPUX and IRIX using the gnu compilers.
However, in order to do so I had changed lines
#ifndef __GNUG__
into
#ifndef __GNUC__
in the files ./src/nco/nco_var_rth.c and
./src/nco/nco_var_scv.c.
There are also the following two type
declarations around line2 429 and 430 in
./src/nco/ncap_utl.c
extern float fmodf(float,float);
extern float fabsf(float);
which I surrounded by a similar #ifndef __GNUC__
as in the two files above.
Does that make sense?
An additional thing: On Irix, the configure
script (correctly) finds getop and getopt.h
somewhere, but the compiler complains about
struct option not defined later on (e.g., in ncap.c). There is a comment in ncap.c, saying that it is unclear how SGI handles getopt; well, I'd conclude nit really well;-))
For me, undefining HAVE_GETOPT and HAVE_GETOPT_H in config.h after ./configure, but previous to make, seems to have fixed the problem.
Otherwise, thanks for making nco available!!!
Regards from Bracknell in the UK,
Christian
( christian_AT_marquardt.fsnet.co.uk , -or-
christian.marquardt_AT_metoffice.com )
Thanks for the tips. It's good to know the configure
mechanism works well with the GNU compilers on
these systems.
> Does that make sense?
Not to me. __GNUG__ applies only to g++, so the
tokens are only meant to be defined if g++ is not
being used. gcc does not want them on my systems.
Change to __GNUC__ should cause"symbol re-defined"
warnings or something like that. I'm afraid if I make the changes you needed to the main tree,
there will be unacceptable regressions in other
environments. We will refer people who use gcc
on SGI or HPPA to your email if the have problems
though.
Thanks for the report!
Charlie