Re: readlink and symlink prototypes [PATCH]

[Jeff Johnston <jjohnstn at redhat dot com>]

Shaun Jackman wrote:
> 2005/11/1, Jeff Johnston <jjohnstn@redhat.com>:
>
> > Could you explain why you need/want to do this?  The reason these
> > prototypes are not outside the #ifdefs is that these functions don't
> > exist in generic newlib.  Generic newlib is not fully SUSv3-compliant
> > nor POSIX-compliant nor is it intended to be.  X86 linux newlib, for
> > example, which is far more robust, overrides the header files.
> >
> > -- Jeff J.
>
>
> My application is cross compiled for an embedded system (arm-elf) as
> well as for a desktop machine for testing purposes. When compiled for
> the embedded system, the application provides its own implementation
> of readlink/symlink. Providing an implementation for an unimplemented
> libc function is quite straight forward; however, patching newlib's
> header files is not so straight forward.
>
> My preference would be for newlib to provide prototypes for all the
> SUSv3/POSIX functions. I'd much rather get link time errors than
> compile time errors for unimplemented SUSv3/POSIX functions.
>
> Cheers,
> Shaun

After thinking about it I realized that my logic above is wrong.  The 
current sys/unistd.h header file has lots of non-generic-newlib 
prototypes already.  Moving these prototypes plus others outside of the 
checks will ultimately make the header file cleaner.

Patch checked in.

-- Jeff J.