Re: safer way to use gnulib

[Michael Goffioul]

On Sat, Mar 20, 2010 at 10:06 PM, Michael Goffioul
<address@hidden> wrote:
> On Sat, Mar 20, 2010 at 9:21 PM, John W. Eaton <address@hidden> wrote:
>> On 20-Mar-2010, Michael Goffioul wrote:
>>
>> | This does not address the problem I reported, does it?
>>
>> Yes, I think it should because I removed the #undef GNULIB_NAMESPACE
>> directive.  That way, accept should not be defined by the gnulib
>> header.  Please try the patch and let me know if it solves the
>> problem.  If not, then please report more details about the problem.
>> Your original report was too vague for me to do anything but guess,
>> and my best guess about the solution to the problem was the change I
>> posted.
>
> Sorry for the incompleteness of the report. I didn't expect you to
> try to fix it as I think the problem is in gnulib. The "accept" definition
> is coming from unistd.h, whose template looks like this:
>
> #if @GNULIB_GETHOSTNAME@
> /* Get all possible declarations of gethostname().  */
> # if @UNISTD_H_HAVE_WINSOCK2_H@
> #  include <winsock2.h>
> #  if !defined _GL_SYS_SOCKET_H
> [snip]
> #   undef accept
> #   define accept               accept_used_without_including_sys_socket_h
> [snip]
> #  endif
> # endif
> #endif
>
> In my case, @GNULIB_GETHOSTNAME@ and @UNISTD_H_HAVE_WINSOCK2_H@
> are both defined to 1. So, I don't really see what will be the effect of
> undefining GNULIB_NAMESPACE on the above.

Some additional info. When scanning the object files, I could
determine that the following files contained redefined "accept"
symbol: defaults.cc, dirfns.cc, file-io.cc, help.cc, input.cc, lex.ll,
oct-hist.cc, oct-procbuf.cc, sighandlers.cc, syscalls, cc, sysdep.cc,
toplev.cc, txt-eng-ft.cc and utils.cc.

Michael.