Fit mask overridden in FoVBackgroundMaker and FluxPointsEstimator

Gammapy version dev

Bug description See:

• https://github.com/gammapy/gammapy/blob/9df7a783fc050cf79667deb94d0cc635a2aa5bfb/gammapy/makers/background/fov.py#L49
• https://github.com/gammapy/gammapy/blob/9df7a783fc050cf79667deb94d0cc635a2aa5bfb/gammapy/estimators/flux_point.py#L906

The dataset.mask_fit if overridden during the computation of flux points or the field of view background normalization. This is bad, because users might want e.g. to:

• normalize the background in a custom energy range
• compute flux points by masking away some map regions

In those cases, they would set a dataset.mask_fit that is silently overridden in the current code. Notice that, for the case of background normalization in a custom energy range, this issue can be overcome by adapting the exclusion mask, instead of touching the dataset.mask_fit. But still, it is legitimate to want to do it with the dataset.mask_fit so it shouldn’t be overridden.

Other information I didn’t check if this issue is also present somewhere else in the gammapy code, it might be worth having a look

[adonath]

@luca-giunti I can’t remember the reasoning either. But looking at this again I think overriding the mask_fit internally is maybe not the best solution. Probably it would make more sense to to keep the mask_fit and combine it internally with the exclusion_mask.

[adonath]

@luca-giunti I think the conclusion from the last dev call here was, to handle everything with the exclusion_mask. Is this correct?