What do you mean by wrap -- do you mean raising a NoDataError if guessing can't happen, or catching it instead of an AttributeError? I skimmed the MDAKits failures and didn't see anything relating to this, although could have missed something due to going fast

@lilyminium looking at some of the failures, MDAnalysis raises a NoDataError not an AttributeError, so the traceback tells you it encountered a NoDataError and then it encountered something else, which is confusing to users. By "wrap", I mean that the except should be looking for both error types.

Sorry, it's late here, I'm not sure I'm fully understanding you -- if I rephrase, the issue is that the traceback is confusing because the Topology raises a NoDataError and the guesser is catching an AttributeError (and then raising a ValueError)? Are you suggesting just switching the guesser to catch a NoDataError? (and IMO raising a NoDataError too as would also fall within the spirit of the error).

@lilyminium I'm saying we should be catching both errors.

Honestly, if this is late on your end then we'll not do the release and wait until these issues are correctly handled.