Re: [PATCH] Workaround for autoconf'ed define of caddr_t

[Esben Haabendal <eha at dev dot doredevelopment dot dk>]

"Yann E. MORIN" <yann.morin.1998@anciens.enib.fr> writes:

> On Wednesday 30 March 2011 16:12:35 Esben Haabendal wrote:
>> (comments in original mail was eaten by gnus)
>
> Yes, indeed. Gnus are special animals, and they feed on comments.
> In this respect, they are a very usefull specy. Trolls also feed on
> comments, but in a very different manner! Fortunately, trolls and
> gnus do not cross-breed. Hehe! ;-)
>
> Anyway, you can still run 'hg email --outgoing' to send changesets in
> your local repos that are not upstream. It even works on MQs! That way,
> I can easily run 'hg import' (or mimport) on your mails to turn the
> patches back into changesets here that I can then push.
>
> And what's nicest with that method, is that it keeps authorship. With
> simple patches as attachements, I have to remember at commit time that
> I am not the author, and force that.

I'll try to remember this next time.

>> When combined with the types.h where caddr_t is protected together with
>> daddr_t, the typedef of caddr_t breaks.
>> 
>> This patch works around it by protecting the caddr_t typedef
>> specifically.
>> 
>> I am uncertain as to the real cause and solution to this :-(
>
> Neither am I... :-/
> I did postponed the issue as I could not understand it as I have more
> important matters going on that I have to handle ASAP.
>
> I'll test your patch before the end of the WE. I'll also look at the glibc
> BT to see if the issue has been reported. Looking at eglibc may also yield
> some hints.

I have not been able to find anything in this direction in glibc
bugzilla nor in gcc bugzilla.

Actually, maybe the "bug" is to be considered to be in gcc
(configure.ac) and not in glibc.  From gcc changelog:

  2008-07-06  Andreas Tobler  <a.tobler@schweiz.org>
    * configure.ac: Check for caddr_t, define to char * if not defined.

Which is implemented in gcc's gcc/configure.ac

  AC_CHECK_TYPE(caddr_t, char *)

Actually, this macro is actually deprecated because of it's flaws
(see http://www.gnu.org/s/hello/manual/autoconf/Obsolete-Macros.html).

But why is AC_CHECK_TYPE not finding the caddr_t type?
I will try to reproduce the problem again, and see if I can find the
reason for this, and hopefully a proper solution (which can go to
upstream gcc or glibc).

/Esben

--
For unsubscribe information see http://sourceware.org/lists.html#faq