Build failure on arm64 with --build-tool=ninja

[thesamesam]

Originally reported downstream in Gentoo at https://bugs.gentoo.org/915544 (w/ regular gcc 13 as well, I just happened to reproduce it w/ gcc 14 below):

$ ./configure.py --build-tool=ninja
   INFO: ./configure.py invoked with options "--build-tool=ninja"
   INFO: Configuring to build Botan 3.2.0 (revision git:6f466a2704a31856ebc27451ca861527d3dd00a1)
   INFO: Python version: "3.11.6 (main, Oct 10 2023, 05:56:40) [GCC 14.0.0 20231010 (experimental) fb124f2a23e92b08556984a50a4b2f367ed04d90]"
   INFO: Autodetected platform information: OS="Linux" machine="aarch64" proc=""
   INFO: Guessing target OS is linux (use --os to set)
   INFO: Guessing to use compiler gcc (use --cc or CXX to set)
   INFO: Guessing target processor is a arm64 (use --cpu to set)
   INFO: Found sphinx-build (use --without-sphinx to disable)
   INFO: Using /etc/ssl/certs/ca-certificates.crt as system certificate store
   INFO: Canonicalized CPU target aarch64 to arm64
   INFO: Auto-detected compiler version gcc 14.0
   INFO: Auto-detected compiler arch arm64
   INFO: Target is gcc:14.0-linux-arm64
   INFO: Assuming target arm64 is little endian
   INFO: Skipping (dependency failure): asio certstor_sqlite3 sessions_sqlite3
   INFO: Skipping (incompatible CPU): aes_ni aes_power8 argon2_avx2 argon2_ssse3 chacha_avx2 chacha_avx512 ghash_vperm idea_sse2 keccak_perm_bmi2 processor_rng rdseed serpent_avx2 serpent_avx512 sha1_sse2 sha1_x86 sha2_32_bmi2 sha2_32_x86 sha2_64_bmi2 shacal2_avx2 shacal2_x86 simd_avx2 simd_avx512 zfec_sse2
   INFO: Skipping (incompatible OS): certstor_system_macos certstor_system_windows commoncrypto win32_stats
   INFO: Skipping (no enabled compression schemes): compression
   INFO: Skipping (requires external dependency): boost bzip2 lzma sqlite3 tpm zlib
   INFO: Loading modules: adler32 aead aes aes_armv8 aes_crystals_xof aes_vperm argon2 argon2fmt aria asn1 auto_rng base base32 base58 base64 bcrypt bcrypt_pbkdf bigint blake2 blake2mac block blowfish camellia cascade cast128 cbc ccm certstor_flatfile certstor_sql certstor_system cfb chacha chacha20poly1305 chacha_rng chacha_simd32 checksum cmac comb4p cpuid crc24 crc32 cryptobox cshake_xof ctr curve25519 des dh dilithium dilithium_aes dilithium_common dl_algo dl_group dlies dsa dyn_load eax ec_group ec_h2c ecc_key ecdh ecdsa ecgdsa ecies eckcdsa ed25519 elgamal eme_oaep eme_pkcs1 eme_raw emsa_pkcs1 emsa_pssr emsa_raw emsa_x931 entropy fd_unix ffi filters fpe_fe1 gcm getentropy ghash ghash_cpu gmac gost_28147 gost_3410 gost_3411 hash hash_id hex hkdf hmac hmac_drbg hotp http_util idea iso9796 kdf kdf1 kdf1_iso18033 kdf2 keccak keccak_perm keypair kmac kuznyechik kyber kyber_90s kyber_common lion locking_allocator mac mce md4 md5 mdx_hash mem_pool mgf1 mode_pad modes mp nist_keywrap noekeon noekeon_simd numbertheory ocb ofb par_hash passhash9 pbes2 pbkdf pbkdf2 pem pgp_s2k pk_pad pkcs11 poly1305 poly_dbl prf_tls prf_x942 psk_db pubkey raw_hash rc4 rfc3394 rfc6979 rmd160 rng roughtime rsa salsa20 scrypt seed serpent serpent_simd sessions_sql sha1 sha1_armv8 sha2_32 sha2_32_armv8 sha2_64 sha3 shacal2 shacal2_armv8 shacal2_simd shake shake_cipher shake_xof simd siphash siv skein sm2 sm3 sm4 sm4_armv8 socket sodium sp800_108 sp800_56a sp800_56c sphincsplus_common sphincsplus_sha2 sphincsplus_shake srp6 stateful_rng stream streebog system_rng thread_utils threefish_512 tls tls12 tls13 tls13_pqc tls_cbc trunc_hash tss twofish utils uuid whirlpool x509 x919_mac xmss xof xts zfec zfec_vperm
   INFO: Using symlink to link files into build dir (use --link-method to change)
   INFO: Botan 3.2.0 (revision git:6f466a2704a31856ebc27451ca861527d3dd00a1) (release dated 20231009) build setup is complete

$ ninja
[9/582] g++ -fPIC -fvisibility=hidden -fstack-protector -pthread -std=c++20 -D_REENTRANT  -O3 -DBOTAN_...build/include/external -c src/lib/block/sm4/sm4_armv8/sm4_armv8.cpp -o build/obj/lib/block_sm4_armv8.o
FAILED: build/obj/lib/block_sm4_armv8.o
g++ -fPIC -fvisibility=hidden -fstack-protector -pthread -std=c++20 -D_REENTRANT  -O3 -DBOTAN_IS_BEING_BUILT -Wall -Wextra -Wpedantic -Wstrict-aliasing -Wcast-align -Wmissing-declarations -Wpointer-arith -Wca
st-qual -Wzero-as-null-pointer-constant -Wnon-virtual-dtor -Wold-style-cast -Wsuggest-override -Wshadow -Wextra-semi  -I build/include -isystem build/include/external -c src/lib/block/sm4/sm4_armv8/sm4_armv8.
cpp -o build/obj/lib/block_sm4_armv8.o
In file included from src/lib/block/sm4/sm4_armv8/sm4_armv8.cpp:8:
/usr/lib/gcc/aarch64-unknown-linux-gnu/14/include/arm_neon.h: In function ‘void Botan::{anonymous}::SM4_E(uint32x4_t&, uint32x4_t&, uint32x4_t&, uint32x4_t&, uint32x4_t)’:
/usr/lib/gcc/aarch64-unknown-linux-gnu/14/include/arm_neon.h:26899:1: error: inlining failed in call to ‘always_inline’ ‘uint32x4_t vsm4eq_u32(uint32x4_t, uint32x4_t)’: target specific option mismatch
26899 | vsm4eq_u32 (uint32x4_t __a, uint32x4_t __b)
      | ^~~~~~~~~~
src/lib/block/sm4/sm4_armv8/sm4_armv8.cpp:38:19: note: called from here
   38 |    B3 = vsm4eq_u32(B3, K);
      |         ~~~~~~~~~~^~~~~~~
/usr/lib/gcc/aarch64-unknown-linux-gnu/14/include/arm_neon.h:26899:1: error: inlining failed in call to ‘always_inline’ ‘uint32x4_t vsm4eq_u32(uint32x4_t, uint32x4_t)’: target specific option mismatch
26899 | vsm4eq_u32 (uint32x4_t __a, uint32x4_t __b)
      | ^~~~~~~~~~
src/lib/block/sm4/sm4_armv8/sm4_armv8.cpp:37:19: note: called from here
   37 |    B2 = vsm4eq_u32(B2, K);
      |         ~~~~~~~~~~^~~~~~~
/usr/lib/gcc/aarch64-unknown-linux-gnu/14/include/arm_neon.h:26899:1: error: inlining failed in call to ‘always_inline’ ‘uint32x4_t vsm4eq_u32(uint32x4_t, uint32x4_t)’: target specific option mismatch
26899 | vsm4eq_u32 (uint32x4_t __a, uint32x4_t __b)
      | ^~~~~~~~~~
src/lib/block/sm4/sm4_armv8/sm4_armv8.cpp:36:19: note: called from here
   36 |    B1 = vsm4eq_u32(B1, K);
      |         ~~~~~~~~~~^~~~~~~
/usr/lib/gcc/aarch64-unknown-linux-gnu/14/include/arm_neon.h:26899:1: error: inlining failed in call to ‘always_inline’ ‘uint32x4_t vsm4eq_u32(uint32x4_t, uint32x4_t)’: target specific option mismatch
26899 | vsm4eq_u32 (uint32x4_t __a, uint32x4_t __b)
      | ^~~~~~~~~~
src/lib/block/sm4/sm4_armv8/sm4_armv8.cpp:35:19: note: called from here
   35 |    B0 = vsm4eq_u32(B0, K);
      |         ~~~~~~~~~~^~~~~~~

I can only reproduce this with --build-tool=ninja (I was keen to try it for the nice progress bars and faster builds). I'll flip us back to make for the ebuild for now.

[randombit]

Thanks for the report - we must not be passing the right flags to enable the instructions here.

BTW in the linked Gentoo bug I noticed you added cpu_use flags for ARMv8 SHA1/SHA2 - here we are using the ARMv8 SM4 instructions and you may want to have a flag for that as well.

[thesamesam]

Thanks @randombit - the PR works for me.

I'll also check out the SM4 bits, we don't currently have a USE flag for it at all (not just in Botan) so I'll file a bug for that.

[securitykernel]

Sorry for the bug, I must have missed that bit. But glad to see that ninja support is being used downstream. 😊