Re: [PATCH] in6_addr: add 64-bit union accessors

[Mike Frysinger <vapier at gentoo dot org>]

On 30 Mar 2015 01:26, Shawn Landden wrote:
> On Sun, Mar 29, 2015 at 08:56:09PM -0400, Mike Frysinger wrote:
> > On 29 Mar 2015 16:15, Shawn Landden wrote:
> > > I'm not going to change the ChangeLog because that practice
> > > breaks git merges.
> > 
> > you still should post it in the commit message.  people don't want to write it 
> > for you.
> > 
> > > --- a/inet/netinet/in.h
> > > +++ b/inet/netinet/in.h
> > > @@ -214,12 +214,14 @@ struct in6_addr
> > >  #ifdef __USE_MISC
> > >  	uint16_t __u6_addr16[8];
> > >  	uint32_t __u6_addr32[4];
> > > +	uint64_t __u6_addr64[2];
> > >  #endif
> > >        } __in6_u;
> > >  #define s6_addr			__in6_u.__u6_addr8
> > >  #ifdef __USE_MISC
> > >  # define s6_addr16		__in6_u.__u6_addr16
> > >  # define s6_addr32		__in6_u.__u6_addr32
> > > +# define s6_addr64		__in6_u.__u6_addr64
> > >  #endif
> > >    };
> > >  #endif /* !__USE_KERNEL_IPV6_DEFS */
> > 
> > in general, i like this, but you're changing the alignment requirements which in 
> > turn means you're possibly changing the ABI.  a library that takes this struct 
> > as an argument built against older glibc would require 32bit alignment, but an 
> > app might be expecting 64bit after this change.  any such impact has to be 
> > considered before merging.
>
> This only applies if glibc is allocating the struct in6_addr, which I am pretty sure
> we do not do.

i mean apps/libs picking up different defs from the glibc headers, not glibc 
itself returning an allocated struct
-mike

Attachment: signature.asc
Description: Digital signature