[core] ximesh files support [terrain type / limit widescan] by sruon · Pull Request #9686 · LandSandBoat/server

sruon · 2026-03-30T18:45:36Z

I affirm:

I understand that if I do not agree to the following points by completing the checkboxes my PR will be ignored.
I understand I should leave resolving conversations to the LandSandBoat team so that reviewers won't miss what was said.
I have read and understood the Contributing Guide and the Code of Conduct.
I have tested my code and the things my code has changed since the last commit in the PR and will test after any later commits.

What does this pull request do?

Credits

I'll start with this because I legitimately can't claim credit for any of this, I am merely the plumber here.

@InoUno for creating the ximeshes files and making them available - also for the excellent xi-visualizer tool from which the bulk of the logic has been lifted
Aamace (not on GitHub), the creator of Xim, which has been useful in figuring out the actual MZB format (ref src/jsMain/kotlin/xim/resource/ZoneDefParser.kt)
@Xenonsmurf for NavMesh-Builder which I referenced for similar purposes
@atom0s for general code snippets here and there and other client decomps relating to client handling of zone data

Description

Wires in a new CZoneMesh class on CZone that loads, decompresses and prepares ximesh data for consumption.
Exposes a general query API to get a cell information starting from world coords and a handful of helpers for terrain and floorId.

There's a lot of data in there which we are not using at the moment, such as barriers but they may come in useful when we swap LOS to use this.
I didn't port the raycast from xi-visualizer at this time.

Note that this introduces zlibstatic as a dependency.
Note that the zone_mesh.cpp code defers to Detour for dtClosestHeightPointTriangle but it is not a hard dependency,

What this enables

Get terrain material at a given world coor.
- Includes an example of restricting Phuabo Tidal Wave to only occur over water.
- This will be key to making scent-based tracking work accurately
Get floor id at a given coord
- Restricts Widescan to only entities on same floor as PC.
- May be useful to start segmenting entities queries by floor in the future?

!pos has been updated to display the information.

Memory

568MB usage for all zones on a single process
The key optimization is the deduplication done at load time to reuse blocks and packed everything tightly.

Perf

As a test I added query in the AI tick path to see how it would cope under future load. About 2us per call in most of my testing.

Loading itself is around 100ms per zone

Steps to test these changes

xi_test has terrain tests.

~~Draft because there's a slight issue with a dozen zone name not matching the ximesh filename, need to fix that first.~~
Added temporary rewriting of mesh filename for the few odd zones. Will work out the real fix in ximeshes repository at some point.

zach2good

Excellent work here! (A million thanks to all the original authors and reference people too!)

zach2good · 2026-03-31T16:36:27Z

+    SOL_REGISTER("getTerrainType", CLuaBaseEntity::getTerrainType);
+    SOL_REGISTER("getFloorId", CLuaBaseEntity::getFloorId);


I'd probably name these currentTerrainType or getCurrentTerrainType etc., or I'd defer to the zone:
zone:getTerrainType(player:getPos()) etc.

Just to really hammer home what's going on - but any of these changes can come in later PRs

Moved to zone bindings.

zach2good · 2026-03-31T16:37:14Z

+        player:setPos(54.84, -27.5, 16)
+        assert(player:getFloorId() == 15)
+    end)
+end)


I'd probably want a couple more cases like for in Ronfaure, where our eventual scent testing will likely be tested - in the river and on the banks being different, etc.

Added a couple positions in East Ronfaure river

zach2good · 2026-03-31T16:38:59Z

    std::unique_ptr<CNavMesh> m_navMesh;
    std::unique_ptr<ZoneLos>  lineOfSight;

+    auto zoneMesh() const -> CZoneMesh*;


Is there any scenario where a zonemesh WONT be populated? Otherwise I'd push to return a reference here instead of a pointer. If the loading can fail or some zones don't have one, then we'd want to return a Maybe, so we enforce the null checking

Is there any scenario where a zonemesh WONT be populated?

Right now we fail gracefully and let the load process continue as there are a couple naming issues we have to resolve but maybe we should back out early if the mesh doesn't load for any reason in the future.

I swapped it to Maybe for the time being.

zach2good · 2026-03-31T16:39:30Z

 find_package(Threads REQUIRED)
 link_libraries(${CMAKE_THREAD_LIBS_INIT})

+CPMAddPackage(


Does this mean we could get rid of our checked-in zlib stuff in the future?

What we have in zlib.cpp isn't actually zlib, it's the FFXI custom Huffman table algorithm

Co-Authored-By: InoUno <774909+InoUno@users.noreply.github.com> Co-Authored-By: atom0s <atom0s@users.noreply.github.com> Co-Authored-By: Mike Murphy <6765305+xenonsmurf@users.noreply.github.com>

zach2good

Bellissimo 👍

sruon force-pushed the navmesh_ground branch 4 times, most recently from 03e1716 to 93f729d Compare March 31, 2026 00:44

sruon marked this pull request as ready for review March 31, 2026 00:46

sruon force-pushed the navmesh_ground branch 2 times, most recently from 054420a to 19a6454 Compare March 31, 2026 07:13

LandSandBoat deleted a comment from github-actions Bot Mar 31, 2026

Xaver-DaRed approved these changes Mar 31, 2026

View reviewed changes

zach2good approved these changes Mar 31, 2026

View reviewed changes

sruon force-pushed the navmesh_ground branch 2 times, most recently from 00a542c to f07e93b Compare April 1, 2026 00:50

Enable ximesh files support

495ccff

Co-Authored-By: InoUno <774909+InoUno@users.noreply.github.com> Co-Authored-By: atom0s <atom0s@users.noreply.github.com> Co-Authored-By: Mike Murphy <6765305+xenonsmurf@users.noreply.github.com>

sruon force-pushed the navmesh_ground branch from f07e93b to 495ccff Compare April 1, 2026 02:17

WinterSolstice8 approved these changes Apr 1, 2026

View reviewed changes

zach2good approved these changes Apr 1, 2026

View reviewed changes

WinterSolstice8 merged commit fb0cb46 into LandSandBoat:base Apr 2, 2026
10 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[core] ximesh files support [terrain type / limit widescan]#9686

[core] ximesh files support [terrain type / limit widescan]#9686
WinterSolstice8 merged 1 commit intoLandSandBoat:basefrom
sruon:navmesh_ground

sruon commented Mar 30, 2026 •

edited

Loading

Uh oh!

zach2good left a comment

Uh oh!

zach2good Mar 31, 2026

Uh oh!

sruon Apr 1, 2026

Uh oh!

zach2good Mar 31, 2026

Uh oh!

sruon Apr 1, 2026

Uh oh!

zach2good Mar 31, 2026

Uh oh!

sruon Apr 1, 2026

Uh oh!

zach2good Mar 31, 2026

Uh oh!

sruon Apr 1, 2026

Uh oh!

zach2good left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

		SOL_REGISTER("getTerrainType", CLuaBaseEntity::getTerrainType);
		SOL_REGISTER("getFloorId", CLuaBaseEntity::getFloorId);

Conversation

sruon commented Mar 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this pull request do?

Credits

Description

What this enables

Memory

Perf

Steps to test these changes

Uh oh!

zach2good left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

zach2good left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

sruon commented Mar 30, 2026 •

edited

Loading