Re: [Hamlib-developer] Building hamlib

[Stephane F. <f8...@fr...>]

On Wed, Jan 29, 2003, Nate Bargmann wrote:
> * Wilbert Knol <zl...@zl...> [2003 Jan 29 06:07 -0600]:
> > 
> > Thanks for reading this. I am having trouble compiling
> > hamlib-1.1.4-cvs-021226
> > 
> > It goes wrong in ./rpcrig at this statement:
> > 
> > [root@zl2bsj rpcrig]# make
> > rm -f rpcrig.h
> > rpcgen -h `test -f rpcrig.x || echo './'`rpcrig.x -o rpcrig.h
> > typedef unsigned hyper setting_x;
> > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> > rpcrig.x, line 42: expected ';'
> > make: *** [rpcrig.h] Error 1
> > [root@zl2bsj rpcrig]#
> 
> It appears the keyword hyper is foreign to C.  Where that is coming from
> I have no idea.  Right now I don't have time to look for it, but I can
> do so this evening if Stephane doesn't answer this.

hyper is an extension to RPC language, which refers to 64 bit long type.

> BTW, Stephane, where are you?  I know he went on a skiing tri[ a few
> weeks back, but I expected to hear something from him by now.  Hopefully
> he is just busy.

I'm back home in Rennes (western France, in Brittany), and I managed to
come back in one piece after my debut at snowboard.hi.
Sorry for the silence, I was a bit busy. I can see the backlog.

> Looking over the message again I see that you invoked make in the rpcrig
> directory.  I'm not sure that this is supported.  I've always run make
> from the root directory of the Hamlib source distribution.  You may want
> to try the later CVS code directly from Sourceforge.  In the file
> README.developer or README.betatester are the preliminary commands
> needed to build the configuration code.  Also included are the required
> versions of the various packages that need to be installed before the 
> autgen.sh script is run.  This may mean some updates to your system.  
> The CVS code of December 26, 2002 had the same requirements as the 
> current CVS.

make in the rpcrig directory is supported, as long as all the targets
from the top directory are already built. Anyway, if you don't see the
difference it may make, it's better to always run make from the root
directory.
In the present case, the CVS version suffers the same as the cvs-021226.
This is a known problem. It looks like some oldish libc-dev distributions
(which rpcgen is part of) don't support hyper keyword. I've seen this
under some version of solaris, linux-ppc and Mac OS X.
The fix (I haven't took the time to implement yet) consists in checking from
a M4 macro called from configure.ac whether rpcgen would accept the
hyper keyword and gcc would compile the result. Failure of this test 
would simply remove rpcrig and rpcrot from the backend list.

More to put on my todo list..

Anyway, in the mean time, you can try to pass --without-rpc-backends to
disable the RPC build. This has just been commited to the CVS
repository.


Cheers,
Stephane