Re: [RFC patch] avoid warning on accesses to hardwired address

[Martin Sebor via Libc-alpha <libc-alpha@sourceware.org>]

On 8/16/21 5:27 AM, Martin Liška wrote:
> Hello.
> 
> Small note here, the patch is incomplete as i386 target still fails:

The Glibc patch I submitted wasn't approved and I haven't implemented
the GCC change yet to handle named address spaces specially.  It's on
my radar but will take some time to resolve.

Martin

> 
> [   96s] cc1: all warnings being treated as errors
> [   96s] make[2]: *** [../o-iterator.mk:9: 
> /home/abuild/rpmbuild/BUILD/glibc-2.34.9000.39.g6e8a0aac2f/cc-base/intl/loadmsgcat.o] 
> Error 1
> [   96s] make[2]: *** Waiting for unfinished jobs....
> [   96s] In file included from ../sysdeps/i386/i686/nptl/tls.h:33,
> [   96s]                  from ../sysdeps/generic/libc-tsd.h:44,
> [   96s]                  from ../include/../locale/localeinfo.h:224,
> [   96s]                  from ../include/ctype.h:26,
> [   96s]                  from loadmsgcat.c:29:
> [   96s] loadmsgcat.c: In function '_nl_load_domain':
> [   96s] ../sysdeps/i386/nptl/tls.h:239:4: error: array subscript 0 is 
> outside array bounds of '__seg_gs struct pthread * __seg_gs[0]' 
> [-Werror=array-bounds]
> [   96s]   239 |   (*(struct pthread *__seg_gs *) offsetof (struct 
> pthread, header.self))
> [   96s]       |   
> ~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> [   96s] ../sysdeps/nptl/libc-lock.h:92:18: note: in expansion of macro 
> 'THREAD_SELF'
> [   96s]    92 |     void *self = 
> THREAD_SELF;                                                 \
> [   96s]       |                  ^~~~~~~~~~~
> [   96s] loadmsgcat.c:770:3: note: in expansion of macro 
> '__libc_lock_lock_recursive'
> [   96s]   770 |   __libc_lock_lock_recursive (lock);
> [   96s]       |   ^~~~~~~~~~~~~~~~~~~~~~~~~~
>