Backend converter bugfix (edge case)

[nkanazawa1989]

Summary

Backend V2 converter raises unexpected error. This is an example code to reproduce the bug.

backend = SomeBackendV1()  # Some backend includes calibration for sx
del backend._properties._gate["sx"]

backend_v2 = BackendV2Converter(backend)
backend_v2.target  # This causes error

TypeError: argument of type 'NoneType' is not iterable

Details and comments

Qiskit Experiments relies on FakeOpenPulse2Q to build own test backend. This backend is defective as it contains calibration for u1, u2, u3, and cx but properties are defined only for u1, u3, cx. This configuration unintentionally reproduces the same situation as written above. The target converter is upgraded to handle this edge case. Missing information in FakeOpenPulse2Q is also added.

[qiskit-bot]

One or more of the the following people are requested to review this:

• @Qiskit/terra-core
• @nkanazawa1989

[coveralls]

Pull Request Test Coverage Report for Build 7592109423

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.04%) to 89.505%

---

Totals
Change from base Build 7586004387:	0.04%
Covered Lines:	59477
Relevant Lines:	66451

---

💛 - Coveralls

[ElePT]

This is an interesting edge case I had not thought about before. I think it's good to have it documented in the test comment and release note. LGTM. Only one question @nkanazawa1989, is this something we would like to backport? thinking especially of the qiskit-experiments dependencies, if we don't backport, this fix will only be available from qiskit 1.0 onwards.

[wshanks]

It would not hurt anything to backport the part adding the extra gates to the FakeOpenPulse2Qproperties, but it is probably not a big deal (it probably isn't used outside of Qiskit and qiskit-experiments?). The qiskit-experiments tests using FakeOpenPulse2Q (converting it to BackendV2) only fail with Qiskit main, not 0.45. I think there have been some other changes to the backend conversion code that triggered the issue, so backporting the fix here would only make sense if backporting the earlier changes.

With this change, the qiskit-experiments tests pass. I was a little worried that they wouldn't because in qiskit-experiments we do a little bit of hacking to make up for the missing property information already and I thought that might conflict with filling in the information here.

[nkanazawa1989]

Thanks @ElePT @wshanks . I'll merge the PR. I don't think the backport is really necessary unless some user writes a bug report issue about this. In practice we don't report calibration without characterizing the gate, so gate property must always exist.

[wshanks]

https://github.com/Mergifyio backport stable/0.46

[mergify]

backport stable/0.46

✅ Backports have been created

• #11788 Backend converter bugfix (edge case) (backport #11609) has been created for branch stable/0.46 but encountered conflicts