v1_1_4 (July 10, 2003)

Notable changes:

- At least some longslit data now is successfully reduced by the
  pipeline.  No guarantees on reduction quality.  Things are probably
  good enough for standard stars, at least.

- New superdarks have been added for 2003 (FCS intensity has been
  greatly decreased).  We do not have any new ones for 900-line data -
  that is needed.  Update your CALIB_DATA files.

- After we implemented the slit-end corrections for v1_1, we figured
  that it'd be OK to exclude only the outermost 4 pixels on each slit,
  rather than the outermost 5.  However, the skylines are actually a
  bit sharper on that 5th pixel than along the rest of the slit; this
  caused the b-spline sky spectrum to be too sharp, especially on
  short slits, and accordingly the sky subtraction to yield
  'hollowed-out', 'donut' residuals (negative in the center of the
  line, positive on the outside).  Rerunning v1_1 reductions is
  probably advisable.

- Shifts between science data and flats are now determined robustly
  when the shifts are >~1.5 pix (via a cross-correlation to get an
  initial guess, followed by the existing centroid method); that was
  not true before.  Wavelength solutions in cases with large shifts
  could definitely be problematic, though..

- In cases where the arc-fitting yields poor results or few lines, we
  now fall back to the optical model + cross-correlation for
  wavelength solution and tilts.

- 'optimal' extractions in EXTRACT1D now uses FWHM=10 pixels if FWHM=0
  is passed to it.  This was a problem in a number of non-DEEP
  reductions, yielding optimal spectra that were 0 everywhere (boxcar
  was still OK).

- wavelength scales on the Allslit files are now correct (was a
  FLOATCOMPRESS issue).