Re: Compile errors with gcc 4.8

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 06/02/2013 20:37, M A Young wrote:
> Fedora rawhide has just moved onto gcc 4.8, and I have had to apply the 
> attached patch to get it to build (I constructed the patch against 4.2.1 
> but I believe it will apply to xen-unstable).
>
> There are two types of problem, the first is from 
> xen/common/compat/memory.c where the error is
>
> memory.c: In function 'compat_memory_op':
> /builddir/build/BUILD/xen-4.2.1/xen/include/public/arch-x86/xen.h:35:33: 
> error: typedef '__guest_handle_const_compat_memory_exchange_t' locally 
> defined but not used [-Werror=unused-local-typedefs]
>       typedef struct { type *p; } __guest_handle_ ## name
>                                   ^
> /builddir/build/BUILD/xen-4.2.1/xen/include/public/arch-x86/xen.h:43:5: 
> note: in expansion of macro '___DEFINE_XEN_GUEST_HANDLE'
>       ___DEFINE_XEN_GUEST_HANDLE(const_##name, const type)
>       ^
> /builddir/build/BUILD/xen-4.2.1/xen/include/public/arch-x86/xen.h:44:41: 
> note: in expansion of macro '__DEFINE_XEN_GUEST_HANDLE'
>   #define DEFINE_XEN_GUEST_HANDLE(name)   __DEFINE_XEN_GUEST_HANDLE(name, 
> name)
>                                           ^
> memory.c:261:13: note: in expansion of macro 'DEFINE_XEN_GUEST_HANDLE'
>               DEFINE_XEN_GUEST_HANDLE(compat_memory_exchange_t);
>               ^
>
> so we are defining something that isn't used.

But it is used, by guest_handle_cast, albeit through several layers of
macros.

I would tentativly object to the use of "__attribute__((unused));" to
work around what appears to be a gcc bug.  If 4.8 support is desperately
wanted, then a better solution would be to specify
-Wno-unused-local-typedefs to disable the buggy feature.

~Andrew

>
> Secondly gcc 4.8 objects to lines like
>
> memset(ctxt, 0, sizeof(ctxt));
>
> where I think you are just zeroing a pointer's worth of memory. There are 
> a few cases of this in the xen code fixed in the patch, and I am also 
> suspicious of line 630 of 
> stubdom/grub-upstream/stage2/fsys_reiserfs.c which is
>
> memset (INFO->blocks, 0, sizeof (INFO->blocks));
>
> which is noted in the build log but not flagged as an error.
>
>  	Michael Young