libgcrypt ECC regression for the use case in GNUNET
Closed, ResolvedPublic
Actions

Assigned To

Authored By

	• gniibe
	Apr 14 2020, 7:18 AM

Description

For some (unknown) reason, GNUNET uses Ed25519 curve with no flags, while it is intended to be used with (flags eddsa) in libgcrypt.

GNUNET's two tests (in src/util) fail with libgcrypt master (to be 1.9.0), while it goes well with 1.8.x.

FAIL: test_crypto_ecdh_ecdsa
...
FAIL: test_crypto_ecc_dlog

rC libgcrypt
	rC0ff36e04f7cd ecc: Remove hard-coded value for ECC_DIALECT_ED25519.

Mentioned Here: rC0ff36e04f7cd: ecc: Remove hard-coded value for ECC_DIALECT_ED25519.

It's a kind of "bug compatibility" but it's a regression anyway.

In the function nist_generate_key (cipher/ecc.c), ec->nbits is number of bits of P.
... while mpi/ec.c sets 256.

make check for src/util of GNUNET now PASS-ed.

PASS: test_crypto_ecdh_ecdsa
...
PASS: test_crypto_ecc_dlog