[Bug] Client crashes on connect whenever a resource registers SCENARIO_POINTS_OVERRIDE_PSO_FILE (FiveM for GTAV Enhanced, early access) — reproducible with a single 10-point file

[Plexus89]

Summary

Any server resource that registers a scenario point manifest override via data_file 'SCENARIO_POINTS_OVERRIDE_PSO_FILE' '.ymt' in its fxmanifest.lua causes every connecting client to crash immediately on FiveM for GTAV Enhanced (early access). Works fine on Legacy.

Unlike the related pool-exhaustion issue for VEHICLE_METADATA_FILE (#162), this crash is not count-dependent — it reproduces with a manifest declaring a single region and a region file containing only 10 scenario points. This doesn't look like a resource-pool limit, it looks like the SCENARIO_POINTS_OVERRIDE_PSO_FILE mounter path is simply broken on Enhanced right now, independent of scale.

Environment

FXServer branch: early-access (Enhanced)
txAdmin: v9.0.0-beta-646aba9a/b110-ea/Win
Client: FiveM for GTAV Enhanced, early access (~2026-08)
OS: Windows 10
Minimal repro

1. data_file 'SCENARIO_POINTS_OVERRIDE_PSO_FILE' 'sp_manifest.ymt' in fxmanifest.lua

2. sp_manifest.ymt declaring exactly one region (downtown)

3. stream/downtown.ymt with 10 valid, in-bounds points copied verbatim from the vanilla Enhanced game files

4. ensure + connect → client crashes right after the last INIT_SESSION callback, no diagnostic info, just [critical] The application has crashed!

5. Comment out just the data_file line, everything else identical → connects fine What we ruled out (in order)

6. Asset format/versioning — ran assets through Alchemist, output byte-identical to input

7. Corrupt source data — used CodeWalker.Core to find 15/144 zone files with points at Position.Z = -21588.934 (~21km below map); traced to mis-converted AerialVehiclePoint entries from the original authoring tool
   Fixed with real reference data — built current CodeWalker master from source for Enhanced NG-key support, generated keys from GTA5_Enhanced.exe, extracted genuine vanilla CScenarioPointRegion files from x64a.rpf, restored every corrupt point 1:1 from vanilla data by XY match. Crash persisted unchanged even with byte-correct vanilla data.

8. Manifest/region mismatch — cross-checked 142-region manifest against physical files, no mismatches

9. Isolated to the mechanism itself — same files present, only the data_file registration toggled → crash toggles with it, 100% reproducible

10. Ruled out pool exhaustion — reduced to 1 declared region / 10 points. Still crashes.

Conclusion
The crash is triggered purely by the presence of a SCENARIO_POINTS_OVERRIDE_PSO_FILE registration, regardless of content, correctness, or scale. This looks unimplemented/broken outright on Enhanced right now rather than a resource-limit issue.

Expected behavior
Resources registering SCENARIO_POINTS_OVERRIDE_PSO_FILE should load without crashing clients, same as Legacy.

Additional notes
Happy to attach the minimal repro resource (1 region, 10 points, vanilla data, ~5KB core files) if useful for the team.

Issue type

Client

Repro rate

Always

Server build version

2026-08

OS

windows10 64Bit

CPU

I7 6700k

GPU

2070Super

RAM

24GB DDR4

Storage type

SSD SATA

Connection type

Ethernet

ISP and bandwidth

No response

DxDiag

No response

Network graph

No response

Platform

Windows

OS version / distribution

No response

CPU

I7 6700k

RAM

24DDR4

Using txAdmin?

None

Hosting provider

None

Machine type

None

/perf endpoint output

No response

DDoS protection

No response

ulimit -n value (Linux only)

No response

Docker Compose file (Docker only)

No response

Steps to Reproduce

1. Create a resource with: fx_version "cerulean" / game "gta5" / data_file 'SCENARIO_POINTS_OVERRIDE_PSO_FILE' 'sp_manifest.ymt'
2. Ship a sp_manifest.ymt (CScenarioPointManifest, PSO) declaring exactly one CScenarioPointRegionDef (e.g. "downtown")
3. Ship a matching stream/downtown.ymt (CScenarioPointRegion, RSC7) containing 10 valid, in-bounds CScenarioPoint entries copied verbatim from the vanilla Enhanced game files
4. ensure the resource on an Enhanced server, connect with an Enhanced client
5. Client crashes right after the last INIT_SESSION callback, no diagnostic info in the client log, just [critical] The application has crashed!
6. Comment out only the data_file 'SCENARIO_POINTS_OVERRIDE_PSO_FILE' line (everything else identical) → client connects fine, no crash

Expected Behavior

Server resources registering a SCENARIO_POINTS_OVERRIDE_PSO_FILE should load without crashing connecting clients, same as on FiveM Legacy.

Actual Behavior

Client crashes immediately on connect ("The application has crashed!") whenever any resource registers a SCENARIO_POINTS_OVERRIDE_PSO_FILE data_file, regardless of file content, correctness, or scale. Reproduces with a minimal file (1 declared region, 10 points, all copied verbatim from vanilla Enhanced game data). Unlike the related VEHICLE_METADATA_FILE pool-exhaustion issue (#162), this is not count-dependent — it isn't a resource-limit being hit, the mounter path itself appears broken/unimplemented on this Enhanced build.

Evidence

minimal_repro_scenario_crash.zip

Additional Context

We spent a long time isolating this on a large (140+ region) real-world scenario-override resource before reducing it to the minimal repro:

1. Asset format/versioning — ran the resource's binary stream assets through Alchemist; output was byte-identical to input, not a format issue.
2. Corrupt source data — used CodeWalker.Core to deserialize every CScenarioPointRegion file and found 15/144 zone files with points at Position.Z = -21588.934 (~21km below map). Traced to mis-converted AerialVehiclePoint entries (type 0, height ~300) from the original authoring tool.
3. Fixed with real reference data — built CodeWalker master from source for Enhanced NG-key support, generated keys from GTA5_Enhanced.exe, extracted genuine vanilla CScenarioPointRegion files from x64a.rpf (levels/gta5/scenario/*.ymt), restored every corrupt point 1:1 from vanilla data by XY match. Crash persisted unchanged even with byte-correct vanilla data.
4. Manifest/region mismatch — cross-checked the 142-region manifest against physical files, no mismatches.
5. Isolated to the mechanism itself — same files present, only the data_file registration toggled → crash toggles with it, 100% reproducible.
6. Ruled out pool exhaustion — reduced to 1 declared region / 10 points. Still crashes.

Conclusion: the crash is triggered purely by the presence of the SCENARIO_POINTS_OVERRIDE_PSO_FILE registration, independent of content/scale. Happy to attach the minimal repro resource (~5KB) if useful.