Closed tangkong closed 1 year ago
Forgot about this PR! Should we wait to tag this repo until lightpath is tagged and available for the CI here?
lightpath is already tagged actually, I'm going to close and re-open this to see what the CI says
I totally forgot I did this too. I'll take another look
fwiw these tests have been passing forever locally, but the experiments-loading bit is confounding travis. This seems to work loading lightpath and its objects if it does exist
the experiments-loading bit is confounding travis
That's annoying but we can ignore it for this PR
I think those test failures can be traced back to:
16:15:28.690 client WARNING MainThread Entry for tst_device_1 is malformed (The information relating to the container class has been modified to the point where the object can not be initialized, please load the corresponding document). Skipping.
16:15:28.706 client WARNING MainThread Entry for tst_device_2 is malformed (The information relating to the container class has been modified to the point where the object can not be initialized, please load the corresponding document). Skipping.
Which implies that the tst db file needs some additional reworking outside of this PR...
Or maybe it means that the tests need happi v2 to pass
No I'm dumb, you're right. The container needs to not rely on lightpath as well, while still working if lightpath exists.
I had switched to LCLSLightpathItem
to be lightpath-compatible, but that may not exist
is lightpath not a dependency of hutch-python? I guess it's an optional? We can just force it to be included in the test suite run
I'll give that a shot then
Ah ok at this point we're waiting for pcdsdevices tag to pull the new LCLSLightpathItem container.
Gotcha, that should be tagged today but I doubt it will make it through the Friday conda-forge pipeline quickly
With https://github.com/pcdshub/pcdsdevices/pull/1065 I'm reverting the containers to LCLSItem, but hopefully leaving the device_class as a SimpleNamespace so that the branches are accessible after being instantiated.
Let's see how fast the tag shows up for pypi
This has turned into a test-suite touchup PR. There's a lightpath bugfix that lets LightController properly handle user supplied configs, but until that is in and tagged some of these tests will be skipped.
This might warrant a quick test run before we merge this, but I've tested loading using a conf.yml that points to an updated happi database
I think this is good to go, right?
Thanks for the merge 👍
Description
Motivation and Context
Closes #349
Should only be merged once other lightpath-related PR's are merged (eg pcdsdevices#1028)
Currently seems to cause a slowdown that I have not isolated. It may be to an increased number of devices being loaded, but testing has not been comprehensive
How Has This Been Tested?
interactively, with quite a bit of pathmunging
Where Has This Been Documented?
This PR