We do a little copy-pasting, and we've been setting dataset and dataset_gcp_project in the default configs.
At best this is harmless, as the default configs are loaded first, then overridden by the analysis-runner defaults, then any manually specified configs.
We do a little copy-pasting, and we've been setting
dataset
anddataset_gcp_project
in the default configs.At best this is harmless, as the default configs are loaded first, then overridden by the analysis-runner defaults, then any manually specified configs.
At worst (if one of these configs is picked manually as a
--config
CLI arg) the dataset/gcp project will be overridden at runtime, leading to errors (https://batch.hail.populationgenomics.org.au/batches/454726/jobs/1 as an example)