Re: [gentooppc-user] First impressions, and a puzzle

[Stefan Boresch]

On Fri, Sep 06, 2002 at 09:07:42AM -0600, Stuart M Jansen wrote:
> *snip*
> > Second, I have a puzzle: The attached C code mimics
> > something used in a routine belonging to the 'babel'
> *snip*

> >   initap=ap;
>
> Not having access to a compiler at the moment and only limited experience with
> variable length C functions, looks like the above line is your problem. Gcc 3.*
> is much more standards compliant than the 2.95 and that _may_ be the reason it
> is rejecting it. If this function truly mimics the real one, then initap isn't
> used and can be removed. The maintainers of babel may not yet have checked to
> make sure their software can compile with the 3.* series. You might want to
> encourage them to.

Thanks for the quick reply (and you spotted the offending line
correctly :-). I should have googled a bit before posting (this problem
has bitten other packages on ppc, debian lists have a lot of stuff:
Anyways, here is the solution (I guess I should file a bug report): 

-------------

>From info libc, variadic function, Argument Macros...:

 - Macro: void __va_copy (va_list DEST, va_list SRC)
     The `__va_copy' macro allows copying of objects of type `va_list'
     even if this is not an integral type.  The argument pointer in
     DEST is initialized to point to the same argument as the pointer
     in SRC.

     This macro is a GNU extension but it will hopefully also be
     available in the next update of the ISO C standard.

If you want to use `__va_copy' you should always be prepared for the
possibility that this macro will not be available.  On architectures
where a simple assignment is invalid, hopefully `__va_copy' _will_ be
available, so one should always write something like this:

     {
       va_list ap, save;
       ...
     #ifdef __va_copy
       __va_copy (save, ap);
     #else
       save = ap;
     #endif
       ...
     }

-------------

Replacing the simple assignment as suggested makes everything
compile...   

Thanks again,

Stefan

-- 
Stefan Boresch
Institute for Theoretical Chemistry and Structural Molecular Biology
University of Vienna, Waehringerstr. 17       A-1090 Vienna, Austria
Phone: -43-1-427752715                        Fax:   -43-1-427752790