Twelve login terminals, root only, console fixes, a faster build, a Niveus splash, and X back on evdev

- tty1-tty12 all run login; the image has only root. /dev/tty7..12 could not
  be written (cdev.c knew minors 10-16 only), which sent their jobs into a
  restart storm that took the supervisor down; fixed, and stat reports them as
  Linux's tty7..12.
- Erasing text at a prompt did nothing: erase_one sent a bare backspace, which
  only moves; it and the echo path now send back-blank-back through vt_feed.
- A white cursor block was left at the end of every line of a verbose boot:
  a line feed at the bottom scrolls the pixels, cursor included, and the
  cursor was erased after the scroll. It comes off first now. A blank cell is
  also marked dirty, which it was not.
- ldlsd died with a general protection fault on real hardware in its variadic
  log function: Linux module code calls with an 8-byte-aligned stack and an
  unset %al, and the prologue's movaps faults. LDLS guest code is built with
  -mstackrealign.
- make iso took over a minute for any change, because staging r8169 rewrote
  the same files and every mtime moved; it now replaces files only if their
  contents differ (about 10 s).
- The loader's splash is redrawn in the boot manager's style, and the boot
  manager shows its cursor whenever a pointer device exists.
- xf86-input-libinput is out of the image again: with it Xorg took both
  keyboard and pointer from libinput and, on the user's laptop, neither worked.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EVRhxi9zpdW2uDUN4FVcao

kernel/dev/vt.c

@@ -1241,20 +1241,26 @@ vt_cycle(int dir)
 void
 vt_feed(struct vt *v, int c)
 {
-  // The grid is the truth, so the old cursor cell can be repainted from it
-  // *after* the byte is consumed -- which lets two cases cost nothing:
-  //  - a byte inside an escape sequence changes nothing on screen;
-  //  - a plain glyph is drawn over the cell the cursor sat on, which repaints
-  //    it anyway (unless a wrap is pending, when it lands on the next line).
+  // A plain glyph is drawn over the cell the cursor sits on, which repaints it
+  // anyway (unless a wrap is pending, when it lands on the next line), so the
+  // cursor needs no separate erase there.
   int mid = v->esc_state != VE_NORMAL;
   int over = !mid && !v->wrap_pending && c >= 0x20 && c < 0x7f &&
              v->u8_need == 0;
-  vt_feed_1(v, c);
-  if (mid && v->esc_state != VE_NORMAL)
-    return;
-  if (over)
+
+  if (over) {
+    vt_feed_1(v, c);
     v->curs_drawn = 0;
-  else
-    cursor_erase(v);
+    cursor_draw(v);
+    return;
+  }
+
+  // Anything else may move the cursor, and a line feed at the bottom scrolls
+  // the *pixels*, block and all. So the cursor comes off first. Erased after
+  // the scroll, it was repainted where the cell had been and the old block
+  // went up a row with the text: a white block at the end of every line of a
+  // verbose boot.
+  cursor_erase(v);
+  vt_feed_1(v, c);
   cursor_draw(v);
 }