Missing storage format tests

[yanesca]

Summary

When I merged Mbed-TLS/mbedtls-framework#83 wrongly assumed that the interface stability failures on the CI can't be related to the framework changes. It turns out that thousands of interface stability tests have disappeared.

We need to confirm that they only have been renamed/moved if that is the case or add them back if not.

See:
https://mbedtls.trustedfirmware.org/blue/organizations/jenkins/mbed-tls-pr-merge/detail/PR-9841-merge/4/pipeline
https://mbedtls.trustedfirmware.org/blue/organizations/jenkins/mbed-tls-pr-merge/detail/PR-9842-merge/4/pipeline
https://mbedtls.trustedfirmware.org/blue/organizations/jenkins/mbed-tls-pr-merge/detail/PR-9835-merge/7/pipeline

[gilles-peskine-arm]

One of the effects of Mbed-TLS/mbedtls-framework#83 was

Fix some test cases that were not properly detected as never-executed.

Sorry, I wasn't very explicit, but in more detail, the script that generates test cases tries to avoid generating positive test cases for mechanisms that are declared, but not implemented. This was buggy, and is more accurate after Mbed-TLS/mbedtls-framework#83. The PR description has an overview of the expected changes:

• Mbed-TLS/mbedtls-framework@119b3b9 — removes some test cases from consideration by abi_check.py, because they would never be executed, because they depend on a mechanism (key type or algorithm) that we don't implement.
• Mbed-TLS/mbedtls-framework@e3f9189 — only changes depends_on lines, which abi_check.py ignores.
• Mbed-TLS/mbedtls-framework@e1f38eb — only changes depends_on lines, which abi_check.py ignores.

Reviewers of Mbed-TLS/mbedtls-framework#83 should have validated those claims. Therefore we expect the interface stability tests to report the changes in 119b3b9740b1a64d0ee8ca302a5dcd72b6d36d3d in consuming PR (#9841, #9796, Mbed-TLS/TF-PSA-Crypto#122).

Looking at the logs for 3.6 and for development, I believe that indeed, the logs are as expected: all the complaints are about disappearing positive key storage tests that use a mechanism that is not implemented. (There's a large number of them because this includes several hashes and several elliptic curves, and we test (almost) the whole 3D matrix of ECDSA-variant × curve × hash.)

I don't know why Mbed-TLS/TF-PSA-Crypto#122 has a passing interface stability test. I'm assuming that this is because the repo is still very new and there's no point of comparison yet? If that's not the problem then maybe the interface stability tests aren't set up properly on TF-PSA-Crypto.

[gilles-peskine-arm]

(I'm not going to close this issue since it's really a review or gatekeeping consideration on Mbed-TLS/mbedtls-framework#83, and I have a conflict of interest as the author of Mbed-TLS/mbedtls-framework#83.)