Re: Update: [pkg-shadow-Bugs][314271] Shadow FTBFS with musl libc

[Rich Felker <dalias@aerifal.cx>]

On Tue, Aug 06, 2013 at 06:03:59PM -0700, Isaac wrote:
> 
> Rich, can we make sizeof(struct in_addr) visible?
> That's ostensibly the only thing necessary for shadow to build at present.
> 
> ----- Forwarded message from pkg-shadow-bugs@alioth.debian.org -----
> 
> Date: Wed, 07 Aug 2013 00:23:43 +0000
> From: pkg-shadow-bugs@alioth.debian.org
> To: noreply@alioth.debian.org
> Subject: [pkg-shadow-Bugs][314271] Shadow FTBFS with musl libc
> 
> pkg-shadow-Bugs item #314271 was changed at 07/08/2013 02:23 by Nicolas Fran??ois
> You can respond by visiting: 
> https://alioth.debian.org/tracker/?func=detail&atid=411478&aid=314271&group_id=30580
> 
> >Status: Closed
> Priority: 3
> Submitted By: Isaac Dunham (idunham-guest)
> Assigned to: Nobody (None)
> Summary: Shadow FTBFS with musl libc 
> Category: None
> Group: None
> >Resolution: Fixed
> 
> 
> Initial Comment:
> I attempted to build shadow 4.1.5.1 with musl libc (http://www.musl-libc.org/), and ran into a few issues:
> 1: missing <sys/socket.h> in libmisc/utmp.c
> glibc includes several headers from <netdb.h>; <sys/socket.h> is one of these.
> 2: libmisc/utmp.c assumes that member sin_addr of
> struct sockaddr_in (type struct in_addr) is completely defined.
> 
> musl has a policy of not making implementation-specific details
> public unless necessary; this is the full definition:

I think you're misunderstanding the issue. It sounds to me like the
program is just failing to include <netinet/in.h>. If struct
sockaddr_in is referenced by one of the musl headers but not defined,
that's a bug in musl, but it sounds to me like the program is just
assuming header pollution (i.e. assuming that some other header
includes netinet/in.h or otherwise exposes this type.

Could you clarify the error that's happening? If it is a bug on musl's
end, I'll fix it. And since this may just involve the utmp.h header
which is not subject to standards requirements, I might work around
the problem on our end even if this is not a musl bug.

Rich