Open sbailey opened 1 month ago
I confirm the issue, and I think it was introduced when we disallowed calibfinder to run with darks in DESI_SPECTRO_CALIB for normal analyses. So I guess the PR above will fix this behaviour by allowing a fallback when run as part of ccdcalib.
To create a new bias with desi_compute_bias, you first have to
unset DESI_SPECTRO_DARK
, or otherwise it fails due to not finding the bias file that you are trying to create in the first place.I did not find this procedure documented, but I have a vague memory of @julienguy mentioning needing to do this. We should fix this so that "missing" biases and darks are only fatal if the processing is actually needs that bias/dark.