Repository navigation
Conversation
…NavIgnore gamedata - sdk -> alliedmodders/hl2sdk@5f891c9 (2026-07-09, fixes g_bUpdateStringTokenDatabase ABI break) - BotNavIgnore linux signature + patch payload synced with Source2ZE/CS2Fixes upstream, validated against live 1.41.6.9 libserver.so
…p clip)
Since the CS2 Season 5 jump rework (buffered jump presses via
CCSPlayerLegacyJump / m_flJumpPressedTime), a mousewheel +jump press
exists only in the subtick move entries. Stripping those entries made
scroll jumps come out clipped ("glued to the floor") while held space
still worked, because the held button survives in the tick-boundary
button state. Keep stripping WASD/duck subtick moves and subtick
viewangles for the consistent tick feel, but let jump through.
|
@M-archand can check and merge this pls |
|
@Andrey309a i forked it, check my profile |
Works ty |
|
@Sharpyku this metamod plugin has issue i can't use mouse scroll jump it keep getting stuck in the air ..... |
Hey, yeah, I know about that issue. I fixed it a long time ago, just go ahead and download the latest fork https://github.com/Sharpyku/STFixes-metamod/actions/runs/29127302797 |
I think this version works better https://github.com/edgegamers/poor-sharptimer |
Okay, if that works better for you, use that one, what you gave me is basically the actual recompiled Sharptimer script, I don’t really see the point but anyway |
|
@Sharpyku After new update its no longer working |
yeah, after the latest update everything got fucked up... i’ll update it again when i have time |
…41.8.2
Metamod:Source 2.0 (plugin API 18, build git1472) dropped SourceHook, so the
previous build could not be loaded at all. Merge SudoGool's port from
SudoGool/STFixes-metamod@metamod-api18, which moves the four hooks to KHook,
bumps the SDK submodule to dadd5d7 and drops the obsolete tier1.a from the link
inputs. Their branch is taken as-is for the submodule pin, since the port was
validated against that SDK.
Two conflicts with this fork's own history, both kept on our side:
- AMBuilder: both sides added common/valveextensions.proto to the protoc
sources and the merge duplicated the line. Removed the duplicate.
- src/detours.cpp keeps dfde32d, the IN_JUMP exemption from subtick input
stripping. Without it a mousewheel jump is eaten on Season 5.
CheckJumpButtonWater no longer resolved after the CS2 update of 2026-09-22
(PatchVersion 1.41.8.2). It is the one lookup that aborts the whole plugin load,
which is why the plugin has been absent from the servers and why the looping
replay bot stopped spawning: BotNavIgnore, the patch that lets bots be added on
maps with no navigation mesh, never got applied.
Valve moved the addressing from [r13+disp32] to [r12+disp32], so 41 C7 85 became
41 C7 84 24 and each mov grew by a byte, which knocked the old wildcard mask out
of alignment. Verified against the live libserver.so: the old pattern matches 0
times, the new one matches exactly once, at 0x15b1cb7.
Audited all 29 linux signatures against that binary. Everything the plugin
requires at load now resolves: the three RESOLVE_SIG entries, the three detours
and BotNavIgnore. ServerMovementUnlock is still broken but is never looked up,
because cs2f_movement_unlocker_enable defaults to 0.
The API 18 port came with a new Windows workflow that copies Metamod's KHook headers into place, but the existing Linux CI was never touched, so it failed on the first translation unit: mmsource-2.0/core/ISmmPlugin.h:489:10: fatal error: 'khook.hpp' file not found core/ISmmPlugin.h includes "khook.hpp" as a sibling of itself, while Metamod ships KHook under third_party/khook/include. Copy the header and its directory next to ISmmPlugin.h before configuring, and fail loudly if they are not where we expect rather than letting the compiler report it several steps later. Also pin the Metamod checkout to 05c5c63 instead of tracking master. That is the build the port was validated against, and the plugin ABI moves without notice; tracking master means a green build today and an unloadable binary tomorrow.
actions/checkout does not resolve an abbreviated SHA. Given ref: 05c5c63 it
treats the value as a ref name pattern and fetches refs/heads/05c5c63* and
refs/tags/05c5c63*, matches nothing, and fails after three attempts:
git -c protocol.version=2 fetch --no-tags --prune --depth=1 origin
+refs/heads/05c5c63*:refs/remotes/origin/05c5c63* ...
The process '/usr/bin/git' failed with exit code 1
Spell the commit out in full. Same commit, same intent as the previous change.
Sharpyku
changed the title
|
@MasterTishere updated |
Since the CS2 update of 2026-10-22 (PatchVersion 1.41.9.2), every server with this plugin crashed about 18 seconds after boot, the moment the first client (a bot or a player) connected: [CS2Fixes] Hook_ClientActive(0, 0, "Dabisi", 0) Segmentation fault (core dumped) The core dump puts the fault in CPlayerManager::OnClientConnected+34, reading gpGlobals->maxClients with gpGlobals == NULL. The StartupServer hook did run (g_pEntitySystem is set in the same function), so it is IVEngineServer2::GetServerGlobals that now returns null under the SDK's vtable layout. Read the globals from INetworkGameServer::GetGlobals instead, which is how CounterStrikeSharp and upstream CS2Fixes get them, and which still works on the updated servers. Also refuse to track a client, rather than crash, if the globals are still unknown when it connects, and make GetPlayer return null in that case.
What
Two separate breaks, either of which is fatal on its own.
Metamod:Source 2.0 dropped SourceHook (plugin API 18, git1469+). The old binary cannot be loaded by it. Ported to KHook:
ClientActive,StartupServer,CTriggerGravity::PrecacheandCTriggerGravity::EndTouchnow go throughKHook::Virtual, thesdksubmodule moves to alliedmodders/hl2sdk@dadd5d7, and the obsoletetier1.ais dropped from the link inputs.This port is @SudoGool's work, from Port STFixes to Metamod API 18 #9. It is merged here unchanged, with their Windows workflow intact, because Port STFixes to Metamod API 18 #9 and this PR cannot both apply cleanly and the two breaks need to ship together.
CheckJumpButtonWaterno longer resolves after the CS2 update of 2026-09-22 (PatchVersion 1.41.8.2,ServerVersion 2000914). It is the one lookup that aborts the whole plugin load, so while it is broken STFixes is simply absent — and with itBotNavIgnore, which is why looping replay bots stopped spawning on surf and bhop maps.Valve moved the addressing from
[r13+disp32]to[r12+disp32], so41 C7 85became41 C7 84 24and eachmovgrew by a byte, knocking the old wildcard mask out of alignment.Linux only; the
windowsline is untouched.Two CI fixes were needed on top of #9, which only added a Windows workflow and left the Linux job untouched:
core/ISmmPlugin.hincludeskhook.hppas a sibling of itself, while Metamod ships KHook underthird_party/khook/include. Without staging it intocore/the Linux build dies on the first translation unit withfatal error: 'khook.hpp' file not found.actions/checkoutdoes not resolve an abbreviated SHA. Givenref: 05c5c63it fetchesrefs/heads/05c5c63*, matches nothing and fails. Metamod is now pinned by full SHA rather than trackingmaster, since the plugin ABI moves without notice.Validation
Linux CI green.
All 29 linux signatures scanned against a live 1.41.8.2
libserver.so. Everything the plugin requires at load resolves single-match: the threeRESOLVE_SIGentries, the three detours, andBotNavIgnore.Known stale, none of them load-blocking:
SetGroundEntitymatches twice — that is upstream CS2Fixes' own signature and ianlucas/cs2-signatures tracks it as ambiguous too, it is not introduced here.ServerMovementUnlockdoes not resolve but is never looked up whilecs2f_movement_unlocker_enableis0.CEntityInstance_AcceptInput,CEntityIdentity_AcceptInput,CBaseEntity_TakeDamageOld,CNavMesh_GetNearestNavAreaandCCSPlayer_WeaponServices_CanUseare stale and unreferenced by any code in this repo.Running on three production Linux servers (two surf, one bhop, steamrt containers) with no crashes, alongside Vauff's Movement Unlocker and GameFixes_mm. Startup log:
Confirmed in game: the looping server-record replay bot is back in the scoreboard on maps that have no navigation mesh, which is what
BotNavIgnoreexists for.Note for anyone running GameFixes_mm
Set its
rampbugandwater_jumpmodules to"enable" "0".FixWaterFloorJumprewrites a float constant in place, so applying both leaves a wrong value.