Armagetron Advanced

Armagetron Advanced 0.2.9.3 is Debian’s package with two patches of ours (fullscreen at the desktop’s resolution by default, a fixed game-controller scheme) and runs with the right picture and sound. Live AI rounds play at 83 steady fps in the benchmark; its played session (2026-10-25) had a 104 fps median, quit cleanly and was fully playable with the pad.

Packages and patches

armagetronad, armagetronad-common (Debian’s 0.2.9.3.0-1 with our patches, packages/armagetronad). Its textures are RGBA8, and the big-endian risk (GL_RGBA or GL_BGRA picked from the SDL surface’s red mask) didn’t bite.

• 0001, fullscreen by default. The compiled-in default was already fullscreen at the desktop’s resolution (read at run time, so a 720p TV gets 720p). The 640x480 window came from lastSuccess, the first fallback when a start fails or is cut short before the display is up: it began as a 640x480 window, and a fresh configuration saved that. It now starts as the default. Only defaults change: a saved mode in the player’s user.cfg is read over them, and the file is never edited.
• 0002, game controllers: a fixed scheme next to the keyboard, nothing bound by hand and nothing written to the configuration. Controller N plays local player N (up to 4): d-pad left/right or L1/R1 turn (one turn per press), R2 or Cross brake (held), the right stick glances, Triangle switches the camera, Select shows the score table (held), Start opens the game menu; in menus the d-pad moves and changes, Cross selects, Circle and Start go back. The buttons are read through SDL2’s game controller API (a DualShock 3 is “PS3 Controller” there): the game is SDL 1.2 on sdl12-compat, whose SDL2 is already loaded, and uGamepad.cpp looks the dozen functions up with dlopen. The pads become key events at the input queue, so recordings and the keyboard are unaffected. Hot-plug: rescanned every second. Settings > Miscellaneous > Game Controllers shows the scheme.
• The quit: the played session quit cleanly once sdl2-compat’s close deadlock was fixed (our sdl2-compat).

How it performs

• Bench (batch 2026-10-18-armagetron-live-ai, one run): 83 steady, 1% low 24.1, one 127 ms frame in 2 minutes, 58 MB free, 28 MB of VRAM, 289 underruns. Not comparable with the playback numbers before it.
• Played: 104 fps median, 1% low 47, 0.3 s of stutter a minute.

The benchmark (recipe): a local game of 4 AI players with the player spectating, round after round (SPECTATOR_MODE_1 and the SP_* settings), fps from the HUD window (WINDOW_STOP=1). On a first start 0.2.9 shows a language menu, player setup and welcome screen that wait for a key; the live variants write FIRST_USE 0 and the language, and during() types Return, Down, Return into a virtual keyboard: “Play Game”, then “Local Game”. The rounds are alike but not identical (the AIs decide live). Until 2026-10-18, base played back assets/armagetronad-bench.aarec (--playback --benchmark), a 90 s recording cut by the record run’s stop: after its end the game draws nothing, so the window held about 6 s of frames. That is variant playback now. A live repeat is in the game’s todo.

Known problems

• On a first start the game shows a language menu and a player setup that wait for a key press.

Investigations

• 2026-10-17 the lost CPU/GPU overlap: Mesa ps3.262’s -13% and the adaptive submission (Mesa 0027).

Research

What the surveys and plans said about it, newest first; each line names its source, the section and the date.

• 2026-10-15, the user’s first wish list (backlog.txt, commit 7d58e3d), “2. GPU driver and game support”: a strong candidate: very light GL, good for long-running stability tests.