If you run kawari-navimesh and give it a zone ID, it can now show
you the loaded collision meshes. I only tested it in inn rooms so
far, but already works super well.
Attempt to improve the accuracy of:
-ItemOperation: Now sends back ItemAcknowledgeAck as well as InventorySlotDiscard and *Fin when discarding items
-ClientTrigger::EventRelatedUnk -> send back commonly observed ActorControlSelf responses
This doesn't make cutscenes play *yet*, but stops the client from
getting stuck waiting for a cutscene to begin. Now the event script
continues, and ends the event.
I re-implemented Unscrambler, but in reverse! This currently only
affects names in the PlayerSpawn packet, it needs to be extended
into others to be considered complete.
See #9
* Lobby server: send back more accurate error messages for problems
* Send the dreaded 2002 error code instead. Exact error codes would probably be better.
-It is unknown if there's a list somewhere.
-Includes an extra check for trying to bypass the client-side
-Update dependencies
-Include a message that selling isn't supported yet
-Display a message indicating an item was bought
We aren't going to be adding quests ever or anytime soon, so all you
can do right now is force every quest to unlock with the aptly named
!completeallquests.
Document some opcodes related to shops and implement a generic gil shopkeeper script
* You can now interact with shopkeepers, and if you have enough gil, you can attempt to purchase items
* Upon trying to buy items the event will auto-cancel for now, because we're missing implementations of several opcodes related to inventory management
This isn't sent in the PlayerStatus packet *yet*, only on
UpdateClassInfo. For some reason, passing it makes the client freak
out and I'm not sure why yet.
This removes a ton of implementation overlap between the two
command systems. For example, we had two different implementations
of unlocking aetherytes which is just unnecessary.
On the flipside, this makes implementing new GM commands just as
easy as writing debug ones. I moved the existing debug Lua
implementations into their GM counterparts and updated the USAGE
accordingly.
This reads the unknown packet data, just throwing it in a byte
buffer. This allows us to read the rest of the segmenets in
the packet, and fixes parsing.
Note that this still log spams with "wrong size" as we can't
determine the size of the unknown packet ahead of time. This is
easy to fix, but it's not a high priority yet.
Fixes#73
The PlayerStatus struct shifted around recently, so I fixed the offsets
yet again. Unlocks should be persistent now, but this as usual requires
a database wipe, sorry!
I also included some refactors of the !unlock debug command that still
had references to the old unlock_action API.
According to the decompiled Lua code, they sometimes can return
negative numbers (like -1, to indicate an error.) I also improved
the debug display of these in the server log a bit.
