newton and newton cg freeze rather than stopping

[0xAlcibiades]

I've seen a few times while developing that newton and newton cg will hang indefinitely sometimes in the case of non-convergence, pegging the CPU with no output.

[stefan-k]

Thank you for reporting! Could you provide a minimal example to reproduce the issue, please?

[gmanthey]

I think I've run into a similar issue with BFGS. I think the issue may be with the solvers starting a new optimization problem for the linesearch in the next_iter function, without passing on things like max_iters or timeout. So a single step may run essentially forever for more complex problems as the default of u64::MAX is used. I would think those options should be passed on even though the linesearch problem should be easier to solve.

[gmanthey]

In my case, changing the width tolerance of the linesearch also solved the issue, suggesting that this is in fact the root cause.