Fix Keccak dispatch race and cross-backend split by dstebila · Pull Request #2426 · open-quantum-safe/liboqs

dstebila · 2026-05-13T16:42:01Z

Claude Code thinks that this might be the cause of the intermittent SHA3 failures. The suggestion and fix seem plausible to me but I don't have access to an AVX512 machine to test it out.

Claude's report:

Keccak_Dispatch() in xkcp_sha3.c (and the equivalent in xkcp_sha3x4.c) overwrote the global sha3_default_callbacks struct field-by-field on the first SHA3 call when AVX512VL was available. Two problems:

Non-atomic struct copy. The assignment sha3_default_callbacks = sha3_avx512vl_callbacks is a member-wise copy of a struct of function pointers. pthread_once only serialized callers of Keccak_Dispatch itself, not concurrent SHA3 operations on other threads, which could observe a partially-overwritten vtable.

Cross-backend split of a single in-flight call. The very call that triggered dispatch had already entered through the xkcp _inc_init. After the swap, subsequent _inc_absorb / _inc_finalize on the same context routed through the newly-installed AVX512VL table, so one in-flight context was initialized by one backend and absorbed/finalized by another.

This is consistent with the intermittent SHA3-512 failures observed in HQC's hash_g on certain AVX512-capable CI runners but not others.

The fix moves the top-level backend selection out of Keccak_Dispatch and into a one-time init_callbacks() in sha3.c / sha3x4.c, guarded by pthread_once, which swaps the callbacks pointer (a single aligned-pointer store) to sha3_avx512vl_callbacks when AVX512 is present. Keccak_Dispatch still sets the xkcp-internal Keccak_*_ptr function pointers for its own use, but no longer touches the top-level table. OQS_SHA3_set_callbacks now triggers the one-time init before assigning the user's value, so auto-detection cannot later clobber a user-provided table.

Copied from slh dsa commit succesfully Info: Updating KAT for HQC-128 Error: Cannot find location clean in upstream list [{'name': 'ref', 'version': '2025-08-22', 'folder_name': '.', 'compile_opts': '-DHQC_ARCH_REF=1', 'signature_keypair': 'crypto_kem_keypair', 'signature_enc': 'crypto_kem_enc', 'signature_dec': 'crypto_kem_dec', 'sources': 'src/ref/parsing.c src/ref/gf.c src/ref/vector.c src/ref/gf2x.c src/ref/reed_solomon.c src/ref/reed_muller.c src/ref/hqc.c src/ref/vector.h src/ref/hqc-1/reed_solomon.h src/ref/hqc-1/parameters.h src/ref/hqc.h src/ref/parsing.h src/ref/data_structures.h src/ref/gf.h src/ref/gf2x.h'}] Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Running "tests/test_kem HQC-128" results in segmentation fault. Build command: cmake -GNinja \ -DOQS_MINIMAL_BUILD="KEM_hqc_128;KEM_hqc_192;KEM_hqc_256" \ .. ninja Removed <immintrin.h> from ref implementations; replaced fips202.c with OQS_SHA3. liboqs can build, but no test cases, no NIST KAT, no test vectors Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

@v

tests/test_kem "HQC-128" still fails with segmentation fault: Process 3191 stopped * thread #2, stop reason = EXC_BAD_ACCESS (code=1, address=0xc8) frame #0: 0x0000000100035088 test_kem`keccak_inc_squeeze(h="", outlen=16, s=0x0000000000000000, r=136) at xkcp_sha3.c:189:18 186 **************************************************/ 187 static void keccak_inc_squeeze(uint8_t *h, size_t outlen, 188 uint64_t *s, uint32_t r) { -> 189 while (outlen > s[25]) { 190 (*Keccak_ExtractBytes_ptr)(s, h, (unsigned int)(r - s[25]), (unsigned int)s[25]); 191 (*Keccak_Permute_ptr)(s); 192 h += s[25]; Target 0: (test_kem) stopped. * thread #2, stop reason = EXC_BAD_ACCESS (code=1, address=0xc8) * frame #0: 0x0000000100035088 test_kem`keccak_inc_squeeze(h="", outlen=16, s=0x0000000000000000, r=136) at xkcp_sha3.c:189:18 frame #1: 0x0000000100034d3c test_kem`SHA3_shake256_inc_squeeze(output="", outlen=16, state=0x0000000100048298) at xkcp_sha3.c:382:2 frame #2: 0x00000001000358dc test_kem`OQS_SHA3_shake256_inc_squeeze(output="", outlen=16, state=0x0000000100048298) at sha3.c:148:2 frame #3: 0x0000000100007924 test_kem`prng_get_bytes(output="", outlen=16) at symmetric.c:45:5 frame #4: 0x0000000100005a54 test_kem`crypto_kem_enc(c_kem="", K="", ek_kem="'x\U00000015\x88\xf6\U0000000f\xf0\xce\U000000101\a፵e\xfe\x92~\r\xb7sSZ\xc4\U0000000e\xd2\xda\xd9%պ\x8e\xee-\xf9\xe07<@v\n\xe6w\x8d?\xe6B\xb7\xb9X\x8b\xae\x86nr@\xcdr!\xd3\U00000002'+[\U00000011\U0000001f\xb0\x8e\xcc\xd0Zl\xd3\xd8-\U00000001Im\xfft\x9b\x9d]\xa2\xd2Nu0\x83\xce}D$\\\xe8o,\x99\x99D\x8c\xb4\U00000016\xb5\xb5\xbd\xfaĸkA Ao\x94｣Yb\a\xcfa\U0000001dQ:\U00000018\xe5\x8cI\xbd/\xd9U\xcb\xe9\U00000019\xb9JD\xfc\x86^\xf6W\U00000016\xa9xz3G\x95\x97Hs5\xc2\U00000001I\xa3\x9c\xd7\xf2\U0000000f\xf8\U00000005\xeb\x92{\xfc\x9fa\U00000018O\x96\xf8\xa7tH[") at kem.c:123:5 frame #5: 0x000000010000306c test_kem`OQS_KEM_hqc_128_encaps(ciphertext="", shared_secret="", public_key="'x\U00000015\x88\xf6\U0000000f\xf0\xce\U000000101\a፵e\xfe\x92~\r\xb7sSZ\xc4\U0000000e\xd2\xda\xd9%պ\x8e\xee-\xf9\xe07<@v\n\xe6w\x8d?\xe6B\xb7\xb9X\x8b\xae\x86nr@\xcdr!\xd3\U00000002'+[\U00000011\U0000001f\xb0\x8e\xcc\xd0Zl\xd3\xd8-\U00000001Im\xfft\x9b\x9d]\xa2\xd2Nu0\x83\xce}D$\\\xe8o,\x99\x99D\x8c\xb4\U00000016\xb5\xb5\xbd\xfaĸkA Ao\x94｣Yb\a\xcfa\U0000001dQ:\U00000018\xe5\x8cI\xbd/\xd9U\xcb\xe9\U00000019\xb9JD\xfc\x86^\xf6W\U00000016\xa9xz3G\x95\x97Hs5\xc2\U00000001I\xa3\x9c\xd7\xf2\U0000000f\xf8\U00000005\xeb\x92{\xfc\x9fa\U00000018O\x96\xf8\xa7tH[") at kem_hqc_128.c:61:22 frame #6: 0x00000001000374dc test_kem`OQS_KEM_encaps(kem=0x0000000100aed6e0, ciphertext="", shared_secret="", public_key="'x\U00000015\x88\xf6\U0000000f\xf0\xce\U000000101\a፵e\xfe\x92~\r\xb7sSZ\xc4\U0000000e\xd2\xda\xd9%պ\x8e\xee-\xf9\xe07<@v\n\xe6w\x8d?\xe6B\xb7\xb9X\x8b\xae\x86nr@\xcdr!\xd3\U00000002'+[\U00000011\U0000001f\xb0\x8e\xcc\xd0Zl\xd3\xd8-\U00000001Im\xfft\x9b\x9d]\xa2\xd2Nu0\x83\xce}D$\\\xe8o,\x99\x99D\x8c\xb4\U00000016\xb5\xb5\xbd\xfaĸkA Ao\x94｣Yb\a\xcfa\U0000001dQ:\U00000018\xe5\x8cI\xbd/\xd9U\xcb\xe9\U00000019\xb9JD\xfc\x86^\xf6W\U00000016\xa9xz3G\x95\x97Hs5\xc2\U00000001I\xa3\x9c\xd7\xf2\U0000000f\xf8\U00000005\xeb\x92{\xfc\x9fa\U00000018O\x96\xf8\xa7tH[") at kem.c:653:10 frame #7: 0x00000001000014fc test_kem`kem_test_correctness(method_name="HQC-128", derand=false) at test_kem.c:253:8 frame #8: 0x0000000100000e94 test_kem`test_wrapper(arg=0x000000016fdfe3b8) at test_kem.c:378:11 frame #9: 0x000000018d86bc58 libsystem_pthread.dylib`_pthread_start + 136 Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

======================================= short test summary info ======================================== FAILED tests/test_binary.py::test_namespace - AssertionError: assert 129 == 0 FAILED tests/test_code_conventions.py::test_style - AssertionError: Got unexpected return code 255 FAILED tests/test_code_conventions.py::test_spdx - assert False FAILED tests/test_kat.py::test_kem[HQC-128] - AssertionError: assert 'xxx' == '1706aa30e811...1fa4b654b2b8e' FAILED tests/test_kat.py::test_kem[HQC-192] - AssertionError: assert 'xxx' == '5e64abaac8f9...e8e610db31188' FAILED tests/test_kat.py::test_kem[HQC-256] - AssertionError: assert 'xxx' == '6d0c22e3c0f0...a8ff5ce8df327' FAILED tests/test_kat_all.py::test_kem[HQC-128] - AssertionError: assert '2d566f0a2743...c6b425f86a116' == 'f604e7edaa4c...007ce066fcd75' FAILED tests/test_kat_all.py::test_kem[HQC-192] - AssertionError: assert '1b0137144f71...39ee740cbbc83' == '4063a6079ded...543c4acd5cb80' FAILED tests/test_kat_all.py::test_kem[HQC-256] - AssertionError: assert 'c13dcd7fa9ec...12232a6fde81e' == '7ee833200a2e...0128d28202ca3' ======================= 9 failed, 274 passed, 4431 skipped in 147.34s (0:02:27) ======================= Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Also: add newline at the end of kats.json Now KATs all pass for HQC-1/3/5. Need to work on namespacing next. Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

For the following symbols, either change to static or add prefix . > nm -g build/lib/liboqs.a Non-namespaced symbol: _code_decode Non-namespaced symbol: _code_encode Non-namespaced symbol: _fft Non-namespaced symbol: _fft_retrieve_error_poly Non-namespaced symbol: _gf_carryless_mul Non-namespaced symbol: _gf_generate Non-namespaced symbol: _gf_inverse Non-namespaced symbol: _gf_mul Non-namespaced symbol: _gf_square Non-namespaced symbol: _vect_mul Non-namespaced symbol: _hqc_pke_decrypt Non-namespaced symbol: _hqc_pke_encrypt Non-namespaced symbol: _hqc_pke_keygen Non-namespaced symbol: _PQCHQC_C_crypto_kem_dec Non-namespaced symbol: _PQCHQC_C_crypto_kem_enc Non-namespaced symbol: _PQCHQC_C_crypto_kem_keypair Non-namespaced symbol: _hqc_c_kem_from_string Non-namespaced symbol: _hqc_c_kem_to_string Non-namespaced symbol: _hqc_dk_pke_from_string Non-namespaced symbol: _hqc_ek_pke_from_string Non-namespaced symbol: _encode Non-namespaced symbol: _expand_and_sum Non-namespaced symbol: _find_peaks Non-namespaced symbol: _hadamard Non-namespaced symbol: _reed_muller_decode Non-namespaced symbol: _reed_muller_encode Non-namespaced symbol: _compute_generator_poly Non-namespaced symbol: _reed_solomon_decode Non-namespaced symbol: _reed_solomon_encode Non-namespaced symbol: _prng_get_bytes Non-namespaced symbol: _prng_init Non-namespaced symbol: _xof_get_bytes Non-namespaced symbol: _xof_init Non-namespaced symbol: _vect_add Non-namespaced symbol: _vect_compare Non-namespaced symbol: _vect_generate_random_support1 Non-namespaced symbol: _vect_generate_random_support2 Non-namespaced symbol: _vect_print Non-namespaced symbol: _vect_sample_fixed_weight1 Non-namespaced symbol: _vect_sample_fixed_weight2 Non-namespaced symbol: _vect_set_random Non-namespaced symbol: _vect_truncate Non-namespaced symbol: _vect_write_support_to_vector Non-namespaced symbol: _code_decode Non-namespaced symbol: _code_encode Non-namespaced symbol: _fft Non-namespaced symbol: _fft_retrieve_error_poly Non-namespaced symbol: _gf_carryless_mul Non-namespaced symbol: _gf_generate Non-namespaced symbol: _gf_inverse Non-namespaced symbol: _gf_mul Non-namespaced symbol: _gf_square Non-namespaced symbol: _vect_mul Non-namespaced symbol: _hqc_pke_decrypt Non-namespaced symbol: _hqc_pke_encrypt Non-namespaced symbol: _hqc_pke_keygen Non-namespaced symbol: _PQCHQC_C_crypto_kem_dec Non-namespaced symbol: _PQCHQC_C_crypto_kem_enc Non-namespaced symbol: _PQCHQC_C_crypto_kem_keypair Non-namespaced symbol: _hqc_c_kem_from_string Non-namespaced symbol: _hqc_c_kem_to_string Non-namespaced symbol: _hqc_dk_pke_from_string Non-namespaced symbol: _hqc_ek_pke_from_string Non-namespaced symbol: _encode Non-namespaced symbol: _expand_and_sum Non-namespaced symbol: _find_peaks Non-namespaced symbol: _hadamard Non-namespaced symbol: _reed_muller_decode Non-namespaced symbol: _reed_muller_encode Non-namespaced symbol: _compute_generator_poly Non-namespaced symbol: _reed_solomon_decode Non-namespaced symbol: _reed_solomon_encode Non-namespaced symbol: _prng_get_bytes Non-namespaced symbol: _prng_init Non-namespaced symbol: _xof_get_bytes Non-namespaced symbol: _xof_init Non-namespaced symbol: _vect_add Non-namespaced symbol: _vect_compare Non-namespaced symbol: _vect_generate_random_support1 Non-namespaced symbol: _vect_generate_random_support2 Non-namespaced symbol: _vect_print Non-namespaced symbol: _vect_sample_fixed_weight1 Non-namespaced symbol: _vect_sample_fixed_weight2 Non-namespaced symbol: _vect_set_random Non-namespaced symbol: _vect_truncate Non-namespaced symbol: _vect_write_support_to_vector Non-namespaced symbol: _code_decode Non-namespaced symbol: _code_encode Non-namespaced symbol: _fft Non-namespaced symbol: _fft_retrieve_error_poly Non-namespaced symbol: _gf_carryless_mul Non-namespaced symbol: _gf_generate Non-namespaced symbol: _gf_inverse Non-namespaced symbol: _gf_mul Non-namespaced symbol: _gf_square Non-namespaced symbol: _vect_mul Non-namespaced symbol: _hqc_pke_decrypt Non-namespaced symbol: _hqc_pke_encrypt Non-namespaced symbol: _hqc_pke_keygen Non-namespaced symbol: _PQCHQC_C_crypto_kem_dec Non-namespaced symbol: _PQCHQC_C_crypto_kem_enc Non-namespaced symbol: _PQCHQC_C_crypto_kem_keypair Non-namespaced symbol: _hqc_c_kem_from_string Non-namespaced symbol: _hqc_c_kem_to_string Non-namespaced symbol: _hqc_dk_pke_from_string Non-namespaced symbol: _hqc_ek_pke_from_string Non-namespaced symbol: _encode Non-namespaced symbol: _expand_and_sum Non-namespaced symbol: _find_peaks Non-namespaced symbol: _hadamard Non-namespaced symbol: _reed_muller_decode Non-namespaced symbol: _reed_muller_encode Non-namespaced symbol: _compute_generator_poly Non-namespaced symbol: _reed_solomon_decode Non-namespaced symbol: _reed_solomon_encode Non-namespaced symbol: _prng_get_bytes Non-namespaced symbol: _prng_init Non-namespaced symbol: _xof_get_bytes Non-namespaced symbol: _xof_init Non-namespaced symbol: _vect_add Non-namespaced symbol: _vect_compare Non-namespaced symbol: _vect_generate_random_support1 Non-namespaced symbol: _vect_generate_random_support2 Non-namespaced symbol: _vect_print Non-namespaced symbol: _vect_sample_fixed_weight1 Non-namespaced symbol: _vect_sample_fixed_weight2 Non-namespaced symbol: _vect_set_random Non-namespaced symbol: _vect_truncate Non-namespaced symbol: _vect_write_support_to_vector Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

However, incorrect namespacing may have caused some symbols from HQC-1 to pollute symbols that should have belonged to HQC-3 and HQC-5 Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

…ded tests] Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

…xtended tests] Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

On certain platforms, "unsigned long" is 32-bit wide, so (1UL << x) can overflow if x is greater than 31, such as with HQC-5. Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

WHY WHY WHY what is even the difference???? Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

For liboqs integration, hash_g implementation has been modified There are some unexpected dispatch errors under liboqs/src/common/sha3 that causes intermittent GitHub Action test failures. Due to the difficulty of debugging on GitHub action, I chose this shortcut of replacing incremental API with a single hash over a continuous strip of memory. There might be performance and stack memory usage penalty associated with this approach, but I deemed it acceptable. we need to eventually figure out what happened with Keccak dispatch and restore the usage of incremental API Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

This reverts commit 955e000. Signed-off-by: Douglas Stebila <dstebila@uwaterloo.ca>

Keccak_Dispatch() in xkcp_sha3.c overwrote the global sha3_default_callbacks struct field-by-field on first SHA3 call when AVX512VL was available. Two problems: 1. The struct copy is non-atomic and races against concurrent SHA3 callers on other threads, who can observe a partially-overwritten vtable. 2. The very call that triggered dispatch had already entered through the xkcp _inc_init. After the swap, subsequent _inc_absorb / _inc_finalize on that same context route through the newly-installed AVX512VL table, so one in-flight context is initialized by one backend and absorbed/ finalized by another. Move the top-level backend selection out of Keccak_Dispatch and into a one-time init in sha3.c / sha3x4.c that swaps the 'callbacks' POINTER (single aligned-pointer store) via pthread_once. Keccak_Dispatch still sets the xkcp-internal function pointers (Keccak_*_ptr) for its own use, but no longer touches the top-level table. Signed-off-by: Douglas Stebila <dstebila@uwaterloo.ca>

sha3_avx512vl_callbacks and sha3_x4_avx512vl_callbacks are declared const; storing them in a non-const callbacks pointer required a const-stripping cast that fails under -Werror -Wcast-qual in the fuzzing build. Store the dispatch pointer as const internally; the public API signature is unchanged (non-const input is fine to assign into a const-pointer slot). Try to fix https://github.com/open-quantum-safe/liboqs/actions/runs/25812670070/job/75833444016?pr=2426 Signed-off-by: Douglas Stebila <dstebila@uwaterloo.ca>

Signed-off-by: Douglas Stebila <dstebila@uwaterloo.ca>

coveralls · 2026-05-13T19:15:24Z

Coverage is 82.372% — sha3-dispatch-fix into gyx-hqc-20250822. No base build found for gyx-hqc-20250822.

The partial-fill-and-permute path in the sha3_absorb macro consumed \`capacity\` bytes via keccak_1600_partial_add (which advances arg2 and clobbers %r12) but did not decrement the remaining message-length tracker. The post-permute path at label 3 then treated the original mlen as still pending, causing the bytes already consumed in the partial fill to be 're-absorbed' (reading past the end of the input buffer) into the post-permute state and leaving s[25] off by \`capacity\`. Symptom: SHA3-{256,384,512} digests are wrong whenever an incremental absorb call exactly fills, or crosses, the rate boundary - reproduced via HQC-3 / HQC-5 KAT failures on AVX512 runners (HQC-1 is unaffected because its hash_g input is 65 bytes, which never crosses the 72-byte SHA3-512 rate). The matching shake_absorb macro in the same file already has the correct pattern (subq %r12, arg3 before the partial_add call); this just mirrors it in sha3_absorb. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> Signed-off-by: Douglas Stebila <dstebila@uwaterloo.ca>

xuganyu96 added 30 commits April 23, 2026 22:18

replace default impl key in doc data sheet

b8f5c11

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[skip ci] copy_from_upstream can run but missing KATs and HQC-192/256

b7040d6

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Fixed HQC PRNG domain; renamed HQC-128/192/256 to HQC-1/3/5

378c164

Also: add newline at the end of kats.json Now KATs all pass for HQC-1/3/5. Need to work on namespacing next. Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Partial fix for namespacing

51e8e70

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Fixed namespacing for HQC

67cdaa6

However, incorrect namespacing may have caused some symbols from HQC-1 to pollute symbols that should have belonged to HQC-3 and HQC-5 Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Still some pollution

49c8d88

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Revert namespace fix

2205d60

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Convenience script for building and testing HQC integration

9dd3b3b

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Applied namespace; KAT hash is incorrect

a025dbc

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Revert OQS_fprintBstr to prevent interference with KAT checksum

8c91d54

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[extended tests] Fixed kat and kat_all checksums

984cd29

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Namespace crypto_memset, set shake256_prng_ctx static

42e3452

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[extended tests] empty commits to trigger extended tests

ee4ad55

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[full tests] Document branching-on-secret suppressions

80e4e11

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[full tests] [extended tests] Cleaned up documentation

2c706c5

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

HQC branching-on-secret is not valgrind-validated [full tests] [exten…

9476794

…ded tests] Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Move convenience script to ignored .env file

9d35576

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[skip ci] Fix memory leak in XOF and hash_<g|h|i|j>

1b0d8fa

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Switch HQC PRNG backend to OQS_randombytes with macro [full tests] [e…

fe31007

…xtended tests] Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Print kat_kem output if checksums don't match

947b907

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[skip ci] debug keygen on Windows

2f78e95

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[skip ci] Does 1UL really mean 32-bit on Windows x86?

8e33b18

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[skip ci] disable debug_keygen

5d8ced9

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Force 64-bit type on BITMASK [extended tests] [full tests]

40e55c7

On certain platforms, "unsigned long" is 32-bit wide, so (1UL << x) can overflow if x is greater than 31, such as with HQC-5. Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

Check pull request readiness again [extended tests]

1e3abde

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

xuganyu96 added 5 commits April 23, 2026 22:18

[skip ci] Additional constant-time test suppression

49641cb

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[full tests] [extended tests] debug logging

dced8ac

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[full tests] [extended tests] astyle formatting

a39afe9

WHY WHY WHY what is even the difference???? Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

[extended tests] Switch from personal fork to official repo

5f0fdfc

Signed-off-by: Ganyu (Bruce) Xu <g66xu@uwaterloo.ca>

dstebila requested review from alexrow, baentsch and bhess as code owners May 13, 2026 16:42

dstebila requested review from xuganyu96 and removed request for alexrow May 13, 2026 16:42

dstebila marked this pull request as draft May 13, 2026 16:43

dstebila added 2 commits May 13, 2026 12:43

Revert "[extended tests] Replace incremental SHA3-512 API"

d3f462d

This reverts commit 955e000. Signed-off-by: Douglas Stebila <dstebila@uwaterloo.ca>

dstebila force-pushed the sha3-dispatch-fix branch from 5e23d5d to 6c93505 Compare May 13, 2026 16:44

dstebila added 2 commits May 13, 2026 12:47

Make callbacks pointer const

3927935

Signed-off-by: Douglas Stebila <dstebila@uwaterloo.ca>

dstebila marked this pull request as ready for review May 13, 2026 21:04

dstebila mentioned this pull request May 13, 2026

Update HQC implementation to 2025-08-22 version #2407

Open

xuganyu96 force-pushed the gyx-hqc-20250822 branch 3 times, most recently from a6b1f68 to f7742f4 Compare May 22, 2026 19:01

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix Keccak dispatch race and cross-backend split#2426

Fix Keccak dispatch race and cross-backend split#2426
dstebila wants to merge 40 commits into
gyx-hqc-20250822from
sha3-dispatch-fix

dstebila commented May 13, 2026

Uh oh!

coveralls commented May 13, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

dstebila commented May 13, 2026

Uh oh!

coveralls commented May 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

coveralls commented May 13, 2026 •

edited

Loading