Jim Bosch added a comment - 27/Sep/22 4:51 PM New output looks like this (obtained manually by attempting a bad QG against the ci_hsc repo; the layering of the CLI stuff in ctrl_mpexec makes it hard to get the big picture from the unit tests): lsst.pipe.base.graphBuilder CRITICAL: Initial data ID query returned no rows, so QuantumGraph will be empty. lsst.pipe.base.graphBuilder CRITICAL: To reproduce this query for debugging purposes, run Registry.queryDataIds with these arguments: lsst.pipe.base.graphBuilder CRITICAL: dimensions=[ 'band' , 'instrument' , 'skymap' , 'detector' , 'physical_filter' , 'tract' , 'exposure' , 'patch' , 'visit' ], lsst.pipe.base.graphBuilder CRITICAL: dataId={ 'instrument' : 'HSC' }, lsst.pipe.base.graphBuilder CRITICAL: where= "skymap='discrete/ci_hsc' AND tract=1 AND patch=69" , lsst.pipe.base.graphBuilder CRITICAL: datasets=[ 'raw' , 'skyMap' , 'jointcalPhotoCalibCatalog' , 'jointcalSkyWcsCatalog' ], lsst.pipe.base.graphBuilder CRITICAL: collections=[ 'HSC/raw/all' , 'HSC/external' , 'refcats' , 'HSC/calib' , 'skymaps' , 'HSC/masks' ], Error: QuantumGraph was empty; diagnostic logging above may have details. What's new here: First line that provides context to the actual diagnostics that follow. log level is CRITICAL, rather than WARNING Final line used to be a RuntimeError with a long traceback, obscuring the important logs. In particular, the middle is unchanged; this an example of a case where the diagnostics can't actually tell you the problem, so they just give you a way to minimally reproduce it.

Jim Bosch added a comment - 28/Sep/22 2:01 PM John Parejko , this is something you requested a while back.  Changes are split across  pipe_base and ctrl_mpexec , and an example of the new output is in a comment above. In both packages there's an extra unrelated commit to deal with some typing issues that popped up due to a new MyPy appearing on pypi since the last merge to main.  

John Parejko added a comment - 29/Sep/22 11:29 PM I don't remember asking for this, but I like it!