Conversation
The texture streaming revert on 1.6 (93e6ae8) restored an older CModelInfoSA.cpp and brought back m_wOtherTimeModel = 0 for requested timed-object models, undoing multitheftauto#4782. CBaseModelInfo::SetColModel then propagates every owned collision assignment to model 0, so the player ped model ends up with an object collision whose data is unloaded, and CPed::ProcessEntityCollision reads a null CColData at 0x5E26EE. Set the pairing back to -1 and backport the explicit paired time model collision tracking from multitheftauto#4864.
While chasing the clang-format failure I noticed the code I backported was the same twenty lines written twice, once for the model itself and once for its day/night partner. It is now two small helpers that apply or restore a custom collision on one model, and SetColModel and RestoreColModel just run them for the partner as well. Same behaviour, about half the code, and the partner logic lives in one place instead of being copy pasted into two functions.
The texture streaming revert in May put 65 files back into their old shape, so every PR against release/1.6.0 has been failing the clang format check since then. This is the output of the exact binary the CI uses (clang-format 21.1.7 from utils/clang-format.ps1) over those 65 files. Whitespace and line breaks only, the preprocessor and token stream are unchanged.
The Windows jobs revert the VS2026 switch commit on the fly so the windows-2022 runner can build. That revert conflicts on win-create-projects.bat since the PREMAKE_FILE change landed on top of it, so all three Windows builds have been failing at that step. Replace the revert with the two edits it was meant to produce.
Contributor
Author
|
I just pushed three follow-up commits. First, I cleaned up the time model code by removing some duplicated day/night logic it acts exactly the same but uses half the code now. I also fixed the clang-format and Windows build failures. Neither were actually caused by this PR, they were just old CI issues (the formatter has been broken since May, and the Windows build script was conflicting with itself). I updated the formatting and tweaked the Windows script so the tests will actually pass now. |
clang-format joined instructions inside __asm blocks in CMultiplayerSA.cpp and the win32 build died with 88 errors. master already wraps every asm block in clang-format off/on (multitheftauto#4611), the 1.6 files never got that. Put the 13 asm files back to how they were, wrapped all 422 blocks, ran the formatter again. Every asm line is byte for byte what it was before.
Contributor
Author
|
yay! |
FileEX
approved these changes
Sep 16, 2026
Member
|
Since two checks won't run for the 1.6.0 branch, someone with permission to merge while bypassing the checks will need to merge it, because auto-merge will wait until all checks pass |
Zephkek
added a commit
to Zephkek/mtasa-blue
that referenced
this pull request
Sep 17, 2026
The texture streaming revert in May put 65 files back into their old shape, so every PR against release/1.6.0 fails the clang-format check. This is the output of the binary the CI uses (clang-format 21.1.7 from utils/clang-format.ps1) over those files, with every __asm block wrapped in clang-format off/on first, the way master does since multitheftauto#4611, because the formatter otherwise merges asm lines and breaks the win32 build. Whitespace, line breaks and the guard comments only; every asm line is byte for byte what it was. Same result as the formatting in multitheftauto#5409.
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Crash at
gta_sa.exe+0x1E26EEinCPed::ProcessEntityCollision. The player ped model carries an object collision whose data is not loaded, so the function reads a nullCColData.Requested timed-object models get
m_wOtherTimeModel = 0, andCBaseModelInfo::SetColModelpropagates every owned collision assignment to that model, which is the CJ ped model. The-1from #4782 did land on 1.6 but was undone by the texture streaming revert in 93e6ae8. This puts it back and backports the explicit paired time model tracking from #4864.Fixes #5310
Test plan
Resource from #3342:
engineRequestModel('timed-object'), replace COL, TXD and DFF, create the object, stop the resource, reconnect. No crash on 1.6 anymore. Same for loading timed-object night models with a map loader and spawning as CJ.Checklist