Commit 5bca5b on main reworks how 0-byte model files are handled ("Change where it bails so maybe its less error prone?"), fixes prediction error counting for certain fields and adds icons for the 5 new TF2 maps.

Fixed: commit 09a162 on main fixes the NPC-related crashes, and the issue was closed.

pruf reported an error loading a map with a 0-byte .lmp file on dev.

BREEDNKMs reported that NPCs crash when searching for cover after the hEnemy change. ZehMatt added a repro; drobotk: spawning a Crow crashes; TotallyARussian: metrocops crash, only on dev.

In #next-update, Yunhaohku hit a crash showing "funny number -12"; Evgeny Akabenko said the cause wasn't determined, and Blank said it happened to him nonstop since the day before. Rubat asked for a crash dump and linked the crash reporting guide. Later that day Darth said Yunhaohku's crash dump shows an outdated or edited launcher.dll.

Commit 7b6a89 on main brings the crash fixes (including saves), allows m_hGoalEnt and m_hPhysicsEnt in SetSaveValue and tries to account for info_ladder_dismount inside the ground.

Commit f668ee on the x86-64 beta: "Fix crashes related to recent changes, including saves".

Cookiestrike reported that vbsp, vvis, vrad and studiomdl crash with access violations after the 22 Oct update; gedlunstts-lgtm posted a Crowbar workaround. In #next-update, lavrentievich, VARdanplay_YT and ejjk123 reported Hammer crashing and Hammer++ failing to compile. Closed 23 Oct.

Commit 038c27 on main handles 0-byte model files that were crashing the game; commit 75a5a6 gets Linux32 building again for now.

evilpigeonlol reported that save and map transitions crash with an empty save file on x86-64 (Windows). In #next-update, Ormi asked the same.

In #lua-help, DoctorHugoSource asked about a new error in an unchanged mod; Redox identified an infinite loop and explained that SetPos now triggers Touch, which calls SetPos again, after the change "Make Entity:SetPos run trigger touch events".

In #next-update, Evgeny Akabenko reported a crash on map load after the x86-64 update (CUtlBuffer::PeekGet invalid offset 0), caused by an addon with 0-byte model files left by a download that ran out of disk space.

Cheatoid posted a snippet crashing x64 with mcore. robotboy655 couldn't reproduce; Cheatoid then concluded it's a race condition.

Commit b9fc35 on main fixes a crash with inputs firing from removed triggers, the Hunter NPC's wrong lookup function and lets SetSaveValue set m_hStoredPathTarget; VPC and gmad/gmpublish now build for Linux64.

Battlefieldkille reported a crash when a removed trigger fires StartTouch. robotboy655 fixed it but the input is now discarded (commit b9fc35).

In #next-update, RaphaelIT7 asked whether "Apply that getmetatable() LuaJIT fix someone wanted" was ported to x86-64: his crashes were gone on dev but not on x86-64. Rubat: "if it isn't in the changelog, stands to reason it wasn't yet".

ArtyomChyornyj2035 reported crashes at an identical client.dll address. robotboy655 said it looks related to a particle system with a C_OP_WorldTraceConstraint constraint tracing against a ragdoll, which nobody could reproduce before; a reproduction is needed.

In #next-update, TheSoulClock said the new x86-64 branch crashes his low-end laptop every time he joins a big server; Rubat said crash logs would be needed to do anything about it. Nykez said a Mac user of his started crashing on load, and WinterPhoenix suspected (not yet verified) a render.Capture crash on x86-64 Linux.

Commit 0e9ec4 on the x86-64 beta brings a large batch of fixes: several potential crashes around entity creation and search, a new p2p_force_reliable convar, loading Srcds workshop IDs from a file, fixed normals on many built-in models and new props.