julian
352c643009
Pins the device-layer bugs we kept hand-verifying, as pure byte-stream assertions (no sockets, no hardware): - printer-escpos.test.ts (12): CP852 codepage select; the ë→0x89 / Ë→0xD3 mapping and the em-dash/⚠ ASCII fallbacks (never a stray 0x3f "?"); and the Code128 MODULE WIDTH contract — a short ticket id at width 3, but the ~20-char out-of-window occurrence id at width 2 so it fits the 80mm head (width 3 overflows ~576 dots and the firmware silently aborts the barcode). Plus the QR-and-Code128 dual encoding and the Albanian stamp() format. - printer-routing.test.ts (6): the failover order (booth printer is a fallback for entry tickets; a receipt never prints on the outside dispenser), rank-then-id tiebreak, and printWithFailover walking the order + NoPrinterAvailableError. Wires Vitest into @parking/devices. devices 18/18 green. Claude-Session: https://claude.ai/code/session_01Xcm6ikLgGoCxxHrxtjkk5V