SimH's sim_os_poll_kbd() uses non-blocking reads (VMIN=0, VTIME=0) that create a race condition with automated input on FreeBSD ptys.

1. Boot hangs after tape bootstrap
2. Ctrl-E interrupt reveals PC = 137634
3. Instruction: TSTB (R0) where R0 = 177560 (TTI status register)

1. Disassemble bootstrap code
2. Identify polling loops:
   • 137616: TTO polling (output)
   • 137634: TTI polling (input) ← BLOCKER
   • 137702: TTO polling (echo)

3. Source code analysis (sim_console.c):

      // Lines 1164-1165: Terminal init
      runtty.c_cc[VMIN] = 0;   // instant return
      runtty.c_cc[VTIME] = 0;  // no timeout
   
      // Lines 1209-1218: Keyboard poll
      t_stat sim_os_poll_kbd (void) {
          status = read(0, buf, 1);
          if (status != 1) return SCPE_OK;  // no char
          return (buf[0] | SCPE_KFLAG);
      }
   

All documented successful boots are on Linux, not FreeBSD. The difference may be in terminal/pty handling or SimH build configuration.

Proposed patch to sim_console.c:

// Option 1: Use blocking read with timeout
runtty.c_cc[VMIN] = 1;
runtty.c_cc[VTIME] = 1;  // 0.1s timeout

// Option 2: Add delay in poll
t_stat sim_os_poll_kbd (void) {
    usleep(1000);  // 1ms delay
    status = read(0, buf, 1);
    ...
}

• Update REBUILD-PLAN.md
• File SimH issue if patch works
• Update this experience report

https://github.com/aygp-dr/unix-v4