re: Automated report: NetBSD-current/i386 build failure

[matthew green <mrg%eterna23.net@localhost>]

NetBSD Test Fixture writes:
>     --- aes_via.o ---
>     In file included from /tmp/build/2024.06.15.20.36.38-i386/src/sys/sys/systm.h:648,
>                      from /tmp/build/2024.06.15.20.36.38-i386/src/sys/crypto/aes/arch/x86/aes_via.c:35:
>     /tmp/build/2024.06.15.20.36.38-i386/src/sys/crypto/aes/arch/x86/aes_via.c: In function 'aesvia_cbc_dec':
>     /tmp/build/2024.06.15.20.36.38-i386/src/sys/lib/libkern/libkern.h:394:33: error: array subscript 'unsigned char[16][1]' is partly outside array bounds of
>  'uint8_t[16]' {aka 'unsigned char[16]'} [-Werror=array-bounds]
>       394 | #define memcpy(d, s, l)         __builtin_memcpy(d, s, l)
>           |                                 ^~~~~~~~~~~~~~~~~~~~~~~~~
>     /tmp/build/2024.06.15.20.36.38-i386/src/sys/crypto/aes/arch/x86/aes_via.c:431:25: note: in expansion of macro 'memcpy'
>       431 |                         memcpy(out + nbytes, tmp, 16);
>           |                         ^~~~~~

i don't understand this.  we've had this warning disabled for
aes_via.c for a while now.  see sys/conf/copts.mk:

1.11         (mrg      08-Aug-23): .if ${MACHINE_ARCH} == "x86_64" || ${MACHINE_ARCH} == "i386"
1.11         (mrg      08-Aug-23): COPTS.aes_ni.c+=     ${CC_WNO_STRINGOP_OVERREAD} ${CC_WNO_STRINGOP_OVERFLOW}
1.11         (mrg      08-Aug-23): COPTS.aes_sse2_subr.c+=      ${CC_WNO_ARRAY_BOUNDS}
1.11         (mrg      08-Aug-23): COPTS.aes_ssse3_subr.c+=${CC_WNO_ARRAY_BOUNDS}
1.11         (mrg      08-Aug-23): COPTS.aes_via.c+=    ${CC_WNO_ARRAY_BOUNDS}
1.11         (mrg      08-Aug-23): .endif

my i386 build does not fail.  pgoyette also mentioned elsewhere
that he saw similar iss with aes_ni.c, which also is listed in
the above change...


(i'm still waiting for my clang/llvm amd64 build to finish or
fail -- i suspect it may be a bad header issue, as my build is
past that portion already.)


.mrg.