[Bug sanitizer/59009] libsanitizer merge from upstream r191666 breaks bootstrap on powerpc64-linux and aarch64-linux

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59009

--- Comment #52 from clyon at gcc dot gnu.org ---
The patch I propose is broken, I missed the new warnings in my build log.
I inserted
#ifdef SYSCALL(chown32)
before
 PRE_SYSCALL(chown16)....
in sanitizer_common/sanitizer_common_syscalls.inc
but the SYSCALL is not defined there, resulting in #if 0 like behaviour.

I'm now wondering how can I get the information to know if
#ifdef __NR_<xxx>uid32
i.e. which headers am I allowed to include here?
Can I add
#include <sys/syscall.h> or is it not desirable?

Quoting Arnd:
> To provide a little more background, uid16 support can be further limited to
> these cases:
>
> * arm32 || x86-32 || sparc32 || sh || m68k
> * libc5 || glibc-2.0 || glibc-2.1 || (glibc-2.2 to 2.15 built against
>   linux-2.2 kernel headers)
>
> or in other words, uid16 is never used
>
> * on 64-bit architectures
> * on any glibc built against 2.4 or later kernel headers
> * on glibc-2.16 or later (which drops support for linux-2.2)
>
> I don't know what the respective minimum libc and kernel versions are
> for llvm and gcc, but it's possible that it could be dropped completely,
> but if there is already a dependency on at least linux-2.3.39 and
> glibc-2.2, the uid16 handling could just be removed.

So if inluding <sys/syscall.h> is not OK in
sanitizer_common/sanitizer_common_syscalls.inc and
b/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.cc (to avoid 
CHECK_TYPE_SIZE(__kernel_old_uid_t);
CHECK_TYPE_SIZE(__kernel_old_gid_t);
)
then it looks like the parts we want to #ifdef out should be removed instead.