[gascarcella]

Phase 2 of #101: the launcher's and in-game menu's side of the second pad (phase 1, #102, was the runtime).

What

• Gamepad tab: Player 2 (input.player2): Off / First gamepad / Second gamepad, a line on what the choice does, and the gamepads connected now with the player each is. Gamepads are counted by SDL instance id, as the runtime does; in the in-game menu that's the game's own count. The table's note says both players use this map, and the Hotkeys tab says player 2's gamepad never fires a hotkey.
• Settings file: player2 is written once set and kept once in the file (as video.menu_size). A bad value gets a warning and reads as "off". The in-game menu applies it live (runtime/live.c already re-applies input).
• The Controls tabs fit the in-game menu at 640x480. Before, the Keyboard and Gamepad tables were 309 and 480 px too wide.
  • Their chips now wrap, as the Hotkeys table's do.
  • The buttons' column is as wide as its widest name. Auto-fitted, it was squeezed below the names once the chips wrapped.
  • The note above a table keeps its Reset button beside it only when both fit. Before, a SameLine() after the wrapped note pushed the next line's start past the edge.
• <PREFIX>_PORT_MENU_SCREEN=controls:<tab> (Screens::show_controls_tab), for a picture of one tab.

Gates that ran

• tests/hello_test.py --menu: everything it ran before, plus the in-game menu's Gamepad tab at 640x480 with player 2 set: nothing wider than its window (new; it failed before the layout fixes). All three Controls tabs fit at 640x480.
• Launcher self-test (dw2003 description): 567/567. New checks: input.player2 written, read back, kept once set to off, a bad value warned about, none written for a file without it. The game's parser gets "player2": "second". Picture 8b-Controls-gamepad.png shows the Gamepad tab with two virtual gamepads.
• hello's launcher self-test: 530/530 (its play path is skipped, as before: hello has no disc). The game reading the string the launcher writes is covered by phase 1's runs, which give the game the same value.
• Windows: the launcher cross-built with no diagnostics; its self-test under Wine passes 564/564.
• Not run: dw2003recomp's scripts/test.sh. Nothing under runtime/, psyq/, cmake/ or the contract changed; the changes are in launcher/ (launcher/ui is compiled into the game only with PSXSTACK_MENU).

Notes

• In the self-test picture both virtual pads show the same name: SDL names a gamepad by its mapping, so two look-alike virtual pads share one. Real pads show their own names.

🤖 Generated with Claude Code