x32 regressions

[Mike Frysinger]

On Wednesday 01 May 2013 21:07:18 Dmitry V. Levin wrote:
> On Wed, May 01, 2013 at 08:40:19PM -0400, Mike Frysinger wrote:
> > On Wednesday 01 May 2013 18:49:02 Dmitry V. Levin wrote:
> > > @@ -230,6 +253,12 @@ printflock(struct tcb *tcp, long addr, int getlk)
> > >  	struct flock fl;
> > >  
> > >  #if SUPPORTED_PERSONALITIES > 1
> > > +# ifdef X32
> > > +	if (current_personality == 0) {
> > > +		printflock64(tcp, addr, getlk);
> > > +		return;
> > > +	}
> > > +# endif
> > 
> > hmm, to review the actual patch now, shouldn't we start a new symbol
> > rather than using X32 ?  i'm pretty sure all these issues X32 is running
> > into also affects MIPS's n32 ABI as they're largely the same concept --
> > native 64bit CPU where the ABI uses 32bit longs/pointers but the syscall
> > API is 64bit.
> 
> Any ideas about the name for this new symbol?

many standards use "ILP32" as the name for this mode:
http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/unistd.h.html
http://publib.boulder.ibm.com/infocenter/zos/v1r12/index.jsp?topic=%2Fcom.ibm.zos.r12.cbcpx01%2Fdatatypesize64.htm

gcc/x32 defines __ILP32__, so using "ILP32" for strace (to avoid collisions) 
seems reasonable.  it might be a little confusing since gcc provides similar 
prefixed defines, but i think that ship sailed long ago with I386 and ALPHA and 
friends.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20130501/b66b4f80/attachment.bin>