tree 98e5c0c3608e9720373380b869fa1f016be0d10f
parent 73692b75378604b8ef68e0ae0a2200d3d482c1f0
author Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com> 1658397501 +0200
committer Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com> 1658403286 +0200

Fix dependencies in test_suite_pk

Using VIA_MD_OR_PSA_BASED_ON_USE_PSA was justified by the fact that
until a few commits ago, the test functions here computed hashes using
either MD or PSA, depending on whether USE_PSA was defined (which itself
was justified by the loose reasoning that "PK is USE_PSA territory").

A few commits ago, test code stopped computing hashes because the hash
values became part of the test data. PK itself does not compute hashes.
As a result, VIA_MD_OR_PSA_BASED_ON_USE_PSA is no longer justified.

There are now two kinds of tests:
- those that only rely on hash data (ECDSA, RSA PKCS#1 v1.5) should
depend on VIA_LOWLEVEL_OR_PSA as that is the minimal dependency, hence
the one used for data
- those that were the layer below PK will internally compute a hash (RSA
PKCS#1 v2.1): currently this hash is always computed using MD (on which
MBEDTLS_PKCS1_V21 depends), so legacy dependencies like MBEDTLS_SHA256_C
should be used for now. The previous dependency was morally wrong, it
didn't show in the driver-only tests only because PKCS#1 v2.1 is
disabled in this test for now.

Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>