Fix #5713: Use pathfinder to find closest ship depot

[SamuXarick]

When ships are asked to find the closest depot, the depot that is provided is not always reachable. This patch provides the closest reachable ship depot, by utilizing the pathfinder (Original, NPF or Yapf) first. Only if it fails, it would then be provided by the original method, distance manhattan based.

https://www.tt-forums.net/viewtopic.php?f=33&t=76137

[nielsmh]

I don't like this approach at all. It's far too intrusive and adds far too many special cases to the pathfinding code, which is already very complex. Start over with something simpler that doesn't need to change code in all the pathfinders.

[nielsmh]

An idea for an alternate approach could be to insert the logic in ShipController instead. If after ChooseShipTrack the VF_PATHFINDER_LOST flag is set, and the current order is go to depot (and it's not a scheduled order) then select a different depot and try again. It's probably necessary to keep a list of depots already tried (and failed), but it can just be local to the ship controller function.

Of course that won't handle the case of OPF, since that never signals no path found. I would suggest not doing anything about that for now, but maybe make a separate change that keeps track of how far a ship has sailed since it last completed an order, and compare that to the expected distance, and send a warning to the player if the ship appears to be taking too long.

[nielsmh]

I tried implementing the approach I outlined above, it's probably not complete like this, but it's significantly shorter and touches much less code: nielsmh@fd90477

[TrueBrain]

We recently switched from Jenkins as CI to Azure Pipelines as CI. This means you need to rebase before this Pull Request will pass its checks. Sorry for the troubles!

[SamuXarick]

@nielsmh
I just experimented your approach and it causes stalls during gameplay when there's too many unreachable depots nearby. It calls the pathfinder for everyone of them.

@J0anJosep has already implemented part of what I had in mind for npf, which got into master, and some of the complexity that was on npf.cpp is gone. Unless you mean the other changes...

[nielsmh]

Still think this looks like too much complexity to attempt to improve a few corner cases.

[SamuXarick]

NoNoCAB v5 (O) 1.7.2.zip
Savegame with 5000 ships for testing

[SamuXarick]

NoCAB v499 (Y), 1.6.1-RC1.sav.zip

Code issues resolved

[LordAro]

I'm with @nielsmh on this - this is far too much code to fix this very minor issue. Perhaps @nielsmh's method could be limited to try the first N (where N is small) number of depots before it reverts to the old Manhattan method?

Also there's currently conflicts

[SamuXarick]

Just resolved conflicts

[andythenorth]

This one appears to be stuck, and reviewers are finding it difficult to providing constructive reviews. Also it appears difficult to be confident that the solutions proposed are correct. On this basis I am closing this one. Thanks for contributing!