svn commit: r223262 - in head: cddl/contrib/opensolaris/lib/libdtrace/common contrib/binutils/bfd contrib/binutils/gas contrib/binutils/gas/config contrib/binutils/ld contrib/binutils/opcodes contr...

[Kostik Belousov]

On Sun, Jun 19, 2011 at 10:38:44AM +0200, Simon L. B. Nielsen wrote:
> 
> On 18 Jun 2011, at 22:48, Jilles Tjoelker wrote:
> 
> > On Sat, Jun 18, 2011 at 01:56:33PM +0000, Ben Laurie wrote:
> >> Modified: head/cddl/contrib/opensolaris/lib/libdtrace/common/dt_subr.c
> >> ==============================================================================
> >> --- head/cddl/contrib/opensolaris/lib/libdtrace/common/dt_subr.c	Sat Jun 18 13:54:36 2011	(r223261)
> >> +++ head/cddl/contrib/opensolaris/lib/libdtrace/common/dt_subr.c	Sat Jun 18 13:56:33 2011	(r223262)
> >> @@ -45,6 +45,7 @@
> >> #include <assert.h>
> >> #include <libgen.h>
> >> #include <limits.h>
> >> +#include <stdint.h>
> >> 
> >> #include <dt_impl.h>
> >> 
> >> @@ -811,15 +812,14 @@ dt_basename(char *str)
> >> ulong_t
> >> dt_popc(ulong_t x)
> >> {
> >> -#ifdef _ILP32
> >> +#if defined(_ILP32)
> >> 	x = x - ((x >> 1) & 0x55555555UL);
> >> 	x = (x & 0x33333333UL) + ((x >> 2) & 0x33333333UL);
> >> 	x = (x + (x >> 4)) & 0x0F0F0F0FUL;
> >> 	x = x + (x >> 8);
> >> 	x = x + (x >> 16);
> >> 	return (x & 0x3F);
> >> -#endif
> >> -#ifdef _LP64
> >> +#elif defined(_LP64)
> >> 	x = x - ((x >> 1) & 0x5555555555555555ULL);
> >> 	x = (x & 0x3333333333333333ULL) + ((x >> 2) & 0x3333333333333333ULL);
> >> 	x = (x + (x >> 4)) & 0x0F0F0F0F0F0F0F0FULL;
> >> @@ -827,6 +827,8 @@ dt_popc(ulong_t x)
> >> 	x = x + (x >> 16);
> >> 	x = x + (x >> 32);
> >> 	return (x & 0x7F);
> >> +#else
> >> +# warning need td_popc() implementation
> >> #endif
> >> }
> > 
> > This commit uncovers breakage that had been present for a while. If I
> > compile this on stable/8 i386 for head i386, _ILP32 is not defined and
> > the warning is hit, breaking the build. Apparently, the code had been
> > broken for a while but I do not use dtrace so I would not have noticed.
> > The tinderboxes have now also noticed the problem so it is not something
> > weird about my system.
> 
> What would be the correct (at least temporary) fix?  Replace the _ILP32 / _LP64 checks with specific architecture checks ?
> 
> I can see that on i386 _ILP32  isn't actually defined - but on amd64 _LP64 is... and it still breaks.
> 
> [simon at zaphod:~] ssh ref9-i386.freebsd.org 'cpp -dM < /dev/null | grep LP'
> [simon at zaphod:~] ssh ref9-amd64.freebsd.org 'cpp -dM < /dev/null | grep LP'
> #define __LP64__ 1
> #define _LP64 1
> [simon at zaphod:~] ssh ref8-amd64.freebsd.org 'cpp -dM < /dev/null | grep LP'
> #define __LP64__ 1
> #define _LP64 1
> 
> I must be missing something...
It breaks on amd64 while building compat32 libraries. The build step
uses i386-targeted compiler, which lacks the ILP32 definition.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/svn-src-all/attachments/20110619/171782c8/attachment.pgp