Here's a fix for fax2ps that corrects behaviour for non-Letter paper
sizes. It fixes two problems:
Without scaling (-S) the fax is now centered on the page size specified
with -H and/or -W. Before, fax2ps was using an obscure and practially
useless algorithm to allocate the image relative to Letter sized paper
which sometime sled to useless whitespace on the paper, while at the
same time cutting of the faxes printable area at the opposite border.
Second, scaling now preserves aspect ratio, which makes unusual faxes
(in particular short ones) print properly.
Fixed PS_Lvl2page() code which outputs non-ASCII85 raw
data. Moved test for when to output a line break to
*after* the output of a character. This just serves
to fix an eye-nuisance where the first line of raw
data was one character shorter than subsequent lines.
Added an experimental ASCII85 encoder which can be used
only when there is a single buffer of bytes to be encoded.
This version is much faster at encoding a straight-line
buffer of data because it can avoid alot of the loop
overhead of the byte-by-bye version. To use this version
you need to define EXP_ASCII85ENCODER (experimental ...).
Added bug fix given by Michael Schmidt to PS_Lvl2page()
in which an end-of-data marker ('>') was not being output
when producing non-ASCII85 encoded PostScript Level 2
data.
Fixed PS_Lvl2colorspace() so that it no longer assumes that
a TIFF having more than 2 planes is a CMYK. This routine
no longer looks at the samples per pixel but instead looks
at the "photometric" value. This change allows support of
CMYK TIFFs.
Modified the PostScript L2 imaging loop so as to test if
the input stream is still open before attempting to do a
flushfile on it. This was done because some RIPs close
the stream after doing the image operation.
Got rid of the realloc() being done inside a loop in the
PSRawDataBW() routine. The code now walks through the
byte-size array outside the loop to determine the largest
size memory block that will be needed.
Added "-m" switch to ask tiff2ps to, where possible, use the
"imagemask" operator instead of the "image" operator.
Added the "-i #" switch to allow interpolation to be disabled.
Unrolled a loop or two to improve performance.
