Comment 9 for bug 1834949

1) I will create a patch and request it to be handled in siesta

2) Your last output: NoCodeModification_CompletelyLinear_0Mixing__GST.out is very wrong. A mixing weight of 0 means that you *never* change the DM/H. Hence you will always do the same calculation over and over. :(
However, I am not surprised in this case either since without the code-modification there is an implicit linear mixing of 1. for the first iteration (regardless of your fdf-input) which explains the drastic jump.

All-in-all, the code modification works and will probably be the *new* default somehow.

Thanks for your analysis.