Re: Purpose of _LIBC in resolv/ ?

[Adhemerval Zanella <azanella at linux dot vnet dot ibm dot com>]

Hi,

>From resolv/README I would say to keep the code as close as possible to libbind (the
same way we keep some for gnulib).  I am not following libbind development, from looks
like from http://wiki.netbsd.org/individual-software-releases/netresolv/ NetBSD has taken
ownership and there is not recent release since 2009.

I am not sure if anyone has taken care to keep glibc in sync with libbind, but doing a
quick check using inet_addr.c against libbind-6.0 I wouldn't say so.  In such cases
I think we need to either decide to keep in sync with libbind (and thus re-testing
and review all the changes that has not landed in glibc) or just remove the _LIBC
defines (which for instance are not complete, since they do not guard the _alias macros).
I am personally in favor of second option.

On 03-06-2014 04:46, Petr Spacek wrote:
> Hello list,
>
> Could somebody explain to me what is the purpose of #ifdef _LIBC in resolv/ ?
>
> 1) Why resolver should behave differently when it is compiled into libc and into application?
>
> 2) It seems that code in resolv/ directory cannot be compiled without _LIBC at least from 2008-12-03 (commit 6f5c3117). I hope I read the code correctly:
>
> int
> __res_vinit(res_state statp, int preinit) {
>         FILE *fp;
>         char *cp, **pp;
>         int n;
>         char buf[BUFSIZ];
>         int nserv = 0;    /* number of nameserver records read from file */
> #ifdef _LIBC
>         int nservall = 0; /* number of NS records read, nserv IPv4 only */
> #endif
> // snip
> #ifdef _LIBC
>                 if (MATCH(buf, "nameserver") && nservall < MAXNS) {
> #else
>                 if (MATCH(buf, "nameserver") && nserv < MAXNS) {
> #endif
>                     struct in_addr a;
>
>                     cp = buf + sizeof("nameserver") - 1;
>                     while (*cp == ' ' || *cp == '\t')
>                         cp++;
>                     if ((*cp != '\0') && (*cp != '\n')
>                         && __inet_aton(cp, &a)) {
> // the problem is that nservall is not declared without _LIBC
>                         statp->nsaddr_list[nservall].sin_addr = a;
>                         statp->nsaddr_list[nservall].sin_family = AF_INET;
>                         statp->nsaddr_list[nservall].sin_port =
>                                 htons(NAMESERVER_PORT);
>                         nserv++;
> #ifdef _LIBC
>                         nservall++;
>                     } else {
> // snip
>
>
> I'm going to write a patch for DNSSEC support. Can I remove this _LIBC magic as the first step, before I dive into DNSSEC?
>
> Please explain me the purpose of this if it has to stay there.
>
> Thank you for your time!
>