Add missing - from fluctuation loop

[mandli]

In the shallow water fwave solver the final loop over the waves is missing a negative sign. As is the if statement was never reaching the case where the speed $s_k = 0$ and splitting the wave $\mathcal{Z_k}$ into both $\mathcal{A}^\pm \Delta Q$.

[rjleveque]

I've never used this particular solver, I guess it's similar to the standard GeoClaw solver rpn2_geoclaw.f but without handling dry states? I notice the geoclaw version just uses < 0 and > 0 rather than comparing to +/- 1e-14. I'm not sure if there's any particular advantage to one or the other...

[mandli]

This is used in a sill example in PyClaw. Dry states are looked for but nothing robust by any means. As to the zero issue, I am not aware of anything where doing 1e-14 instead of 0 would be advantageous. If we do keep the non-zero option we of course will need the sign kept.

[ketch]

Git blame says I committed that code 10 years ago, when the Geoclaw solver didn't work with PyClaw. I think it might have originally been written by Matteo Parsani even earlier. I have no idea now why I used that 1e-14 threshold, and the omission of the minus sign definitely seems like an error.

Regarding the threshold, I think that the only really problematic situation is the near-dry-state causing a divide-by-almost-zero in the calculation of beta, but the dry_tolerance parameter already protects against that, so I would favor changing the threshold for s to zero.

[mandli]

Yeah, I would not push the dry-state thing but it is relatively good about handling it for easy cases.