ctools

Using ctselect with the ethres=DEFAULT parameter, results in an observation list with various energy ranges. This works fine.
If one however subsequently runs ctselect with a small energy range, where only few runs contribute with events, an exception is thrown:

*** ERROR in GEbounds::insert_eng(int&, GEnergy&, GEnergy&): Invalid argument. Invalid energy interval specified. Minimum energy 607.552 GeV can not be larger than maximum energy 200 GeV.
*** ERROR encounterted in the execution of ctselect. Run aborted ...

This can be reproduced by running csspec on with a larger energy range than the actual preselected observations.
The problem occurs in ctselect::set_ebounds(). At the end of this function, we create GEbounds with emin and emax. If e.g. emin was adjusted to match previous cuts it may happen that emin>emax leading to the above exception while creating the ebounds.
My suggestion is to add a check when creating the ebounds:

    // Set selection energy boundaries
    GEbounds result;

    if (emin > emax) {
        result.append(GEnergy(emin,"TeV"), GEnergy(emin,"TeV"));
    }
    else {
        result.append(GEnergy(emin, "TeV"), GEnergy(emax, "TeV"));
    }

    // Return result
    return result;
 

Note that in case emin>emax, we could use a GEbounds with [emin, emin]. This circumvents the exception and assures the selection of zero events. This zero energy range also allows to predict zero events, e.g. for ctmodel, etc.

However, on the gammalib-level, an exception is then thrown:

*** ERROR encounterted in the execution of ctlike. Run aborted ...
*** ERROR in GObservation::npred_spec(GModel&, GTime&): Invalid energy range (Emin=0 MeV, Emax=0 MeV) specified.

This can easily be fixed by allowing emin=emax in GObservation::npred_spec(). Accordingly, we could change if(emin<=emax) to if(emin<emax) in GObservation::npred_spec(). I guess allowing npred=0 in case the energy width is 0.0 makes more sense anyway