svn commit: r293903 - in head/sys/boot: arm/uboot efi/fdt efi/libefi

[Steven Hartland]

On 14/01/2016 16:53, Warner Losh wrote:
>
>
> On Thu, Jan 14, 2016 at 2:47 AM, NGie Cooper <yaneurabeya at gmail.com 
> <mailto:yaneurabeya at gmail.com>> wrote:
>
>
>     > On Jan 14, 2016, at 01:27, NGie Cooper <yaneurabeya at gmail.com
>     <mailto:yaneurabeya at gmail.com>> wrote:
>     >
>     >
>     >> On Jan 14, 2016, at 01:22, Steven Hartland <smh at FreeBSD.org> wrote:
>     >>
>     >> Author: smh
>     >> Date: Thu Jan 14 09:22:01 2016
>     >> New Revision: 293903
>     >> URL: https://svnweb.freebsd.org/changeset/base/293903
>     >>
>     >> Log:
>     >> Fix GCC warnings causing build failure after r293724
>     >>
>     >> Disable some compiler warnings for GCC (non-standard compiler)
>     fixing
>     >> build failures introduced by r293724, which enabled WARNS in
>     the EFI boot
>     >> code, when compiling with none standard compiler (GCC).
>     >
>     > Disabling warnings is wrong. Here’s how arm could be fixed:
>     >
>     > $ svn diff sys/boot/
>     > Index: sys/boot/common/bootstrap.h
>     > ===================================================================
>     > --- sys/boot/common/bootstrap.h (revision 293877)
>     > +++ sys/boot/common/bootstrap.h (working copy)
>     > @@ -32,6 +32,7 @@
>     > #include <sys/types.h>
>     > #include <sys/queue.h>
>     > #include <sys/linker_set.h>
>     > +#include <time.h>
>     >
>     > /*
>     >  * Generic device specifier; architecture-dependant
>     > @@ -332,12 +333,8 @@
>     > extern struct arch_switch archsw;
>     >
>     > /* This must be provided by the MD code, but should it be in the
>     archsw? */
>     > -void   delay(int delay);
>     > -
>     > void   dev_cleanup(void);
>     >
>     > -time_t time(time_t *tloc);
>     > -
>     > #ifndef CTASSERT                /* Allow lint to override */
>     > #define CTASSERT(x)             _CTASSERT(x, __LINE__)
>     > #define _CTASSERT(x, y)         __CTASSERT(x, y)
>     > Index: sys/boot/efi/include/efiapi.h
>     > ===================================================================
>     > --- sys/boot/efi/include/efiapi.h       (revision 293877)
>     > +++ sys/boot/efi/include/efiapi.h       (working copy)
>     > @@ -532,6 +532,7 @@
>     > typedef
>     > EFI_STATUS
>     > (EFIAPI *EFI_RESERVED_SERVICE) (
>     > +    VOID
>     >     );
>     >
>     > typedef
>     >
>     > amd64 is a bit trickier:
>     >
>     >
>     /scratch/tmp/ngie/svn/sys/boot/efi/libefi/../include/efigop.h:75:
>     warning: 'ms_abi' attribute directive ignored
>     >
>     > I didn’t dig into why this is happening [yet], other than there
>     maybe being MS extensions involved?
>     >
>     > Thanks,
>     > -NGie
>
>     This might fix amd64:
>
>     Index:
>     /scratch/tmp/ngie/svn/sys/boot/efi/libefi/../include/amd64/efibind.h
>     ===================================================================
>     ---
>     /scratch/tmp/ngie/svn/sys/boot/efi/libefi/../include/amd64/efibind.h
>           (revision 293877)
>     +++
>     /scratch/tmp/ngie/svn/sys/boot/efi/libefi/../include/amd64/efibind.h
>           (working copy)
>     @@ -160,8 +160,10 @@
>      //
>
>      #ifdef __amd64__
>     +#if defined(__clang__) && (__clang_major__ >= 3 &&
>     __clang_minor__ >= 5)
>      #define        EFIAPI  __attribute__((ms_abi))
>      #endif
>     +#endif
>
>      #ifndef EFIAPI                  // Forces EFI calling conventions
>     reguardless of compiler options
>          #ifdef _MSC_EXTENSIONS
>
>
>
> Sure this fix is right? If you aren't marking the ABI correctly, how 
> could this possibly work?
>
> Warner
No I don't think it is, Andrew's fix is the most appropriate (disable 
building of all EFI on none supported platforms).

Currently I intend to backout 293903 and apply this change instead. I'm 
just waiting on a full tinderbox under GCC to finish.

     Regards
     Steve