Failing GATT tests after deprecating Data signing in Zephyr Host

[PavelVPV]

The following tests are failing in Zephyr Host after deprecating Data signing (zephyrproject-rtos/zephyr#99204):

• GATT/SR/GPA/BV-04-C
• GATT/SR/GAR/BV-04-C
• GATT/SR/GAD/BV-04-C
• GATT/SR/GAD/BV-05-C

There was an attempt to update the GATT Database (#88) but it seems that the database is generated from some other source.

Since the source is not clear for me, I've created this issue to follow-up on this later while these tests are commented out in Zephyr.

[aescolar]

CC @MariuszSkamra @sjanc @Tronil @wopu-ot @thoh-ot @niag-Oticon @rugeGerritsen

[wopu-ot]

The GATT Database was originally generated from the "GATT Qualification Test Database" Excel sheet that can be found along with the Core specifications (e.g., https://www.bluetooth.com/specifications/specs/core-specification-6-2/). However, there seem to be some more or less subtle changes and I don't think we could regenerate it easily today.

I am wondering why we would need a change in the data base here. The database provided along with the Core specification does not seem to be dependent on the LE security mode. What exactly breaks without the change? Do our test implementations make assumptions that they should not make?

[PavelVPV]

I am wondering why we would need a change in the data base here. The database provided along with the Core specification does not seem to be dependent on the LE security mode. What exactly breaks without the change? Do our test implementations make assumptions that they should not make?

These tests are accessing a specific service which was using "Authenticated Signed Writes" property. The property is now deprecated in Zephyr Host. Thus, not possible to build Zephyr Host with Data signing feature support.

These tests if I'm not mistaken, don't depend on that property itself, but the service they were accessing was using it. If I remove the property on the DUT, Database Hash mismatches between what Host generates and this auto-generated database.

[wopu-ot]

I think I understand now that the change in #88 is basically matching the change in tests/bsim/bluetooth/ll/edtt/gatt_test_app/src/gatt/service_f_1.c in zephyrproject-rtos/zephyr#99204 (sorry for being a little slow on that).

It would be interesting to know how the official tests handle the case that the reference data base contains items that the DUT does not support.

On the one hand, it is not nice to deviate from the original data base provided by BT SIG. On the other hand, EDTT does not have support for an ICS (see #14), so any changes to the test cases would need to be based on assumptions, which is not nice either. Given the precedence by the Zephyr change, I guess going for the former approach is okay as long as we document the reasoning.

[sjanc]

Typically Lower Tester is PTS not IUT so this is not an issue. Same problem occurs when some Behavior Invalid tests require invalid behavior from Lower Tester, so when IUT is used for LT is may be a problem.

Personally I think Sec Mode 2 is such irrelevant feature that I'd simply alter DB and don't care much about it (and just add a comment where deviations from reference DB is present for clarity).

In long term maybe there should be reference Lower Tester available for EDTT but this is likely not going to happen anytime soon.