Re: ac-archive: ac_sys_largefile and libraries...

[Guido Draheim]

Paul Eggert schrieb:
> > Date: Sat, 11 Jan 2003 11:00:57 +0100
> > From: Guido Draheim <address@hidden>
> >
> > The end of the discussion can only be to change the compilation
> > behaviour globally - making 64bit off_t the default if that is
> > supported by the system, and dropping the settable off_t altogether.
>
>
> I suppose we could change Autoconf so that largefile-mode is the
> default, and you need to invoke a new macro AC_SYS_SMALLFILE if you
> want 32-bit off_t applications on 64-bit off_t hosts.  That wouldn't
> solve the problem in general, but it would solve it for programs that
> use Autoconf.

See the note about zlib - it might screw you up.

To reiterate in the matter - it is okay for an application (!!) to
enable largefile as default, it is even okay for a library to do
so as long as the usage is hidden in the .c sources itself - and
never exported in a header file (e.g. gtk does it right). The
problem only exists when off_t creeps into public headers usable
by third party applications, as most library makers are not
aware that off_t is *not* a fixed type. (I'll leave out applications
that install header files used by "plugins" here, same problem
there, it's in the public headers, included with another config.h).

Well, still I'd like to pick up your idea, just differently - as
to extend AC_TYPE_OFF_T to check whether the local system is a
largefile sensitive one - and issue a hint message on the user
screen and config log. Just a hint, not even a warning or some
such, just changing the answer

-checking for off_t... yes
+checking for off_t... yes (largefile sensitive)

A developer will surely see it and try to get the answer as to
what that actually means - and be more knowledgable afterwards
and ready to avoid the problemspace that is hidden in it.

WDYT, cheers, guido