Closed jl-wynen closed 1 week ago
Strange, I doubt the files have changed.
This PR in scippnexus was merged on september 4th, and the first failing nightly build is on september 5th... So I am guessing that's the culprit?
Maybe. But I don't see how logs (in that PR) and transforms (the failing code) are connected.
From a glance it seems like the attribute vector
did exist on the data-group (that's why it worked before), but not on the data-array, so when we replace the data-group with the data-array we lose the attributes.
It's worth mentioning that the nexus files in this example are broken. Particularly NXlog
don't look like what we expect.
https://github.com/scipp/scippnexus/pull/231/ is indeed the issue, reverting to the last commit without the change works fine with the current notebook.
This is now fixed in scippnexus by being more graceful when encountering broken files.
But to make the nightly pass we need to release scippnexus (~in progress~ done) and lower bound the scippnexus dependency in essreflectometry to that release.
Nightly builds are broken because the Amor files can no longer be loaded. The error occurs in
scippnexus/nxtransformations.py:62
:where
self.attrs.get('vector') is None
.I don't know whether this is a problem with the file or ScippNeXus.