Temporary calibration DB

[alecandido]

I've been to actually run on qw11q, without exporting my own platforms (thus relying on those installed on the system).

[codecov]

Codecov Report

Attention: Patch coverage is 47.82609% with 12 lines in your changes missing coverage. Please review.

Project coverage is 70.24%. Comparing base (0ca412b) to head (e5ce826).

Files	Patch %	Lines
src/qibolab/instruments/qm/controller.py	47.82%	12 Missing ⚠️

Additional details and impacted files

@@                  Coverage Diff                  @@
##           default-threshold     #981      +/-   ##
=====================================================
- Coverage              70.30%   70.24%   -0.07%     
=====================================================
  Files                     64       64              
  Lines                   6679     6701      +22     
=====================================================
+ Hits                    4696     4707      +11     
- Misses                  1983     1994      +11     

Flag	Coverage Δ
unittests	70.24% <47.82%> (-0.07%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[alecandido]

Of course, I have to write the tests, but I already requested the review such that you could also experiment by yourself.

It's an ugly fix, and maybe we should really report this as a bug to QM.

I checked for the repo for the qm-qua package, but apparently is not public (not linked on the PyPI page of the package, and I guess the organization would be https://github.com/qua-platform, where there isn't).

[stavros11]

Thanks @alecandido. I reproduced the issue and confirmed this fixes it.

[stavros11]

Do we really need this? We could always just use the tempfile solution even when we have write access and it is not strictly required. I wouldn't expect a big overhead.

[alecandido]

My only concern is that the temporary file will always be erased upon disconnection. So you will always lose that.

[stavros11]

Sorry, I could not write this earlier. I just wanted to note that this is probably not relevant if we are just executing, as this shouldn't change the content of the file, but could be relevant if we perform the mixer calibration. Currently the driver provides some functions to do that, but I think they are not that useful (I even erased from 0.2). Still, it may indeed be useful to keep this option in case we do that in the future. Another point is that mixer calibration cannot be done without write access, so it will not be possible with the global cluster installation.

[alecandido]

Thanks @stavros11 for the comment.

I agree this option might be useless, but it also has a default, so it's no harm. For the time being, merging it was the conservative approach, but we can revert it later on.

(removing the option would be a breaking change, but it's relevant only if the options will be truly used by someone...)

[scarrazza]

Thanks, also works for me.

[alecandido]

I realized that the tests will be compromised in any case by 0.2, and they should run hardware. I'm merging this and #980 as hotfixes, refined support and tests will be rolled out in 0.2 (to avoid investing too much in something we'll change).