Newton with trust regions questions / issues

[NittanyLion]

Two questions / issues:

1. Line 372 of newton_trust_region.jl has

delta = 0.25 * norm(x_diff)

but this appears to be a new delta. Should the left hand side be state.delta?

2. Various tolerances are hard-coded. For instance, lines 24 and 28 have 1e-10. Are these choices still optimal for big floats?

[pkofod]

delta = 0.25 * norm(x_diff)

Oh no, that is definitely a bug. Thank you!

Various tolerances are hard-coded No, they're not even necessarily optimal for any problem with Float64. They could certainly be exposed.

[NittanyLion]

Thanks!