unit-tests-latest-qiskit fails since mid-Oct

[1ucian0]

Informations

• Qiskit Aer version: dev
• Python version: 3.11
• Operating system: ubuntu-latest in CI

What is the current behavior?

Since Oct 17, there are 24 failing tests when Qiskit dev is installed. There is a chance this is a Qiskit issue that introduced a regression in the stable API, I didn't check it. If so, please open an issue in Qiskit/qiskit

Steps to reproduce the problem

Go to https://github.com/Qiskit/qiskit-aer/actions/workflows/unit-tests-latest-qiskit.yml?page=2 and see the errors.

What is the expected behavior?

Aer should work with Qiskit dev branch, ie. main

[gadial]

This is a well-known problem, that should be fixed in #2265 . Basically, save_expval was extended with a smart truncation technique allowing aer to run on circuits compiled vs many-qubits backends, but the implementation had a small bug causing problems in several edge cases. Should be ok once the PR is merged.

[Cryoris]

Nice! But it would be interesting to know what in Qiskit triggered these failing tests... do you think it has something to do with how you now fixed the Paulis being mapped to the qubits?

[gadial]

I'm not sure what changed on the qiskit side to trigger them. Basically the bug requires a relatively rare situation to arise (truncating lower-indices qubits while keeping higher-indices qubits intact) which can be very dependent on the specifics of the transpilation. I didn't look any deeper than that.