Re: Symbol availability in C, C++

[Paul Eggert]

Albert Chin <address@hidden> writes:

> On Mon, Nov 28, 2005 at 04:43:06PM -0800, Paul Eggert wrote:
>> http://lists.gnu.org/archive/html/bug-gnulib/2005-11/msg00051.html
>
> Well, that certainly solves the uintmax_t redefinition problem.

OK, thanks, I installed it into gnulib.

> But, it doesn't solve the mbstate_t problem on HP-UX.

As I understand it from
<http://lists.gnu.org/archive/html/bug-gnulib/2005-11/msg00085.html>,
the problem is that the HP-UX C compiler by default operates in c89
(or perhaps even K&R?) mode, whereas if you want a C compiler that is
compatible with aCC (your C++ compiler), you're supposed to use the C
compiler in C99 mode.

If that is the issue, how about if you arrange to invoke the HP-UX C
compiler in C99 mode if that is available?  Having C be compatible
with C++ would be a win overall, I'd think.

This could be done by using coreutils/m4/c.m4; see
<http://savannah.gnu.org/cgi-bin/viewcvs/coreutils/coreutils/m4/c.m4>.
We could add c.m4 to gnulib, and require it for modules that have this
sort of problem.  Eventually this c.m4 should migrate into Autoconf so
the problem will go away in the long run.

Perhaps coreutils/m4/c.m4 does not address the particular HP-UX
problem mentioned in the original bug report (I see that c.m4 has no
special entries for HP-UX and C99) but I hope it could be made to do
so.