// the take
A 2-4 player top-down shooter in C/SDL2 with a real authoritative-server netcode stack: client-side prediction, snapshot interpolation, and a shared simulation core used by the local game, the online server, bots, and the test suite. It's aimed at developers who want a small, readable reference for how UDP game netcode actually works, or a group of friends who want a LAN/internet shooter to hack on over a weekend.
The core design is legitimately good: game.c contains all simulation logic with zero SDL dependency, so the exact same code runs in the local build, the headless server, the bots, and the unit tests — no drift between 'what the client thinks happens' and 'what the server says happens'. Fixed 1/60s timestep via accumulator means simulation behavior is decoupled from frame rate. Client-side prediction with input replay against server snapshots, plus ~100ms interpolation for remote players, is the correct approach and rare to see done properly in a project this size. CI actually runs the netcode (two bots playing a full match under sanitizers) rather than just compiling, and there's a real test suite (15 tests under ASan/UBSan) for the rules engine.
Packets are raw little-endian C structs sent directly over the wire with no htonl/ntohl conversion — the README admits this, but it means the protocol is non-portable (breaks on big-endian hardware) and, more practically, there's no validation that a struct cast onto arbitrary network bytes is memory-safe against a malicious or just malformed sender. There's no authentication or encryption on the UDP server, so any client (or script) can spoof input packets or impersonate a player ID. Getting an online match running requires manual IP sharing, router port-forwarding, or self-managed cloud VM firewall rules — there's no matchmaking, NAT traversal, or relay, which is a real barrier for the 'play with friends online' use case the README leads with. Zero stars, zero forks, single contributor, and a Halloween-themed name suggests this is a seasonal hobby project with no signal yet on whether it'll be maintained past October.
We read the README, the directory tree and a key source file in November 2026. Nobody pays to be listed. How we judge →