TURN RECORD: FIRMWARE CONVERSION HELPERS PROBED DURING A REAL EXPORT.

Addresses derived from the CALLA immediates (file = cal - 0x7E40000):
  0x07FC229C -> file 0x18229C (bit 0x3FE114E0)
  0x07FC1EB2 -> file 0x181EB2 (bit 0x3FE0F590)
  0x07FC1F5E -> file 0x181F5E (bit 0x3FE0FAF0)
  0x07FC2210 -> file 0x182210 (bit 0x3FE11080)

Run: ./xie/build/dec/decclient 192.168.99.2:0 ff 3   (the genuine 2.3 client,
     org=BandByPlane, export straight to the window)

MEASURED:
  image: org=3 w=64 h=64 levels=2,0,0 arSize=4096
  photomap created
  put tile sent
  X Error: BadValue, Major 129 (Xie), Minor 21 (Export)

  trace:
    H1EB2 a0=003F09B0 a1=00000000 a8=40C4D134
    H1EB2 a0=003F09B0 a1=00000000 a8=44CFF000

  H229C (0x7FC229C) never fired, H1F5E and H2210 never fired, and none of the
  post-call sites (R_C4 / R_E2 / R_10 / R_44) fired either.

So with org=BandByPlane the export does NOT go through the per-element block at
0x1B3C4..0x1B43E at all -- the only conversion helper reached is 0x7FC1EB2, twice,
with the same source pointer (0x3F09B0) and a1 = 0, and float-looking results in
A8 (0x40C4D134 / 0x44CFF000).  Those are the only two conversion calls observed.

INTERPRETATION (not yet proven): 0x7FC1EB2 is being used as a numeric helper
(its outputs look like IEEE singles), not as the pixel mover; the pixel-moving
call that would fail is not reached, which is consistent with BadValue being
raised before any real conversion.

STATE (unchanged): no client-chosen image displayed; window untouched; white vs
black captures differ only by Terminal Manager figures and cursor.
xshow (core X) remains the only viewer that has shown an image.

STATUS.  NO client-chosen XIE image.
