The CPI file-format

		(C) 1997 Yann Dirson <dirson@debian.org>


 This file documents the CPI font-file-format, as understood by version 0.94
and above of the Linux console utilities ('kbd').

 This file has revision number 1.0, and is dated 1997/09/02.
 Any useful additionnal information on CPI files would be great.


0. Changes

1998/08/20: updated author's e-mail.


1. Summary

 The CPI file format is used by MS-DOS to store its fonts. This description
comes from reverse-ingeneering on codepage(1), so it's very incomplete. Any
complement would be greatly appreciated.

 A CPI file contains fonts for several MS-DOS code-pages, each in possibly
several sizes (typically 08, 14, 16). Each of these fonts is designed for a
particular device, which can be either a screen or a printer device.

 There is also a DR-DOS CPI format, for which I have even less documentation.

 MS-DOS 6.22 provides the following CPI files containing screen (EGA) fonts:

	iso.cpi
	ega.cpi
	ega2.cpi
	ega3.cpi
	
 The 3 latter files just contain selected codepages from the first one, in
sizes 08, 14 and 16, but the first one only contains size 16.


2. History

 Unknown.


3. Known programs understanding this file-format.

 The following program in the Linux console utilities can read and/or write
PSF files:

	codepage (R)


4. Technical data

 The file format is described here in sort-of EBNF notation. Upper-case
WORDS represent terminal symbols, ie. C types; lower-case words represent
non-terminal symbols, ie. symbols defined in terms of other symbols.
 [sym] is an optional symbol
 {sym} is a symbol that can be repeated 0 or more times
 {sym}*N is a symbol that must be repeated N times
 Comments are introduced with a # sign.



# The data (U_SHORT's) are stored in LITTLE_ENDIAN byte order.

ms_cpi_file =
(off = 0)	ms_cpi_file_header
(off = 23)	font_info_header
(off = 25)	cp_entry_header
(off = 53)	<more>				# chained data in no special order

ms_cpi_file_header =
(off = 0)	ms_cpi_id
(off = 8)	font_res	# ?
(off = 16)	num_pointers	# ?
(off = 18)	p_type		# ?
(off = 19)	offset
 (size = 23)
			
ms_cpi_id	= {CHAR}*8 = 0xFF "FONT   "	# cpi_id is a 8-bytes "magic number" string
 (size = 8)
font_res	= {CHAR}*8
 (size = 8)
num_pointers	= U_SHORT
 (size = 2)
p_type		= CHAR
 (size = 1)
offset		= U_LONG
 (size = 4)

font_info_header = nb-codepages
 (size = 2)

nb-codepages = U_SHORT
 (size = 2)

#

# The header's normal size is 28 bytes. However, it is said there happens to
# be 26-bytes headers. What happens in this case ? Is font_offset still valid ?
cp_entry_header =
(off = 0)	cp_header_size 
(off = 2)	next_header_offset 
(off = 6)	device_type
(off = 8)	device_name
(off = 16)	cp_number
(off = 18)	cp_res			# ?
(off = 24)	font_offset
 (size = 28)

cp_header_size		= U_SHORT		# size of the codepage_header
 (size = 2)
next_header_offset	= U_LONG
 (size = 4)
device_type		= U_SHORT = 1 | 2	# 1 = screen ; 2 = printer
 (size = 2)
device_name		= {CHAR}*8		# eg. "EGA     "
 (size = 8)
cp_number		= U_SHORT
 (size = 2)
cp_res			= {CHAR}*6
 (size = 6)
font_offset		= U_LONG
 (size = 4)

<AT FONT-OFFSET>: cp_info_header cp_fontdata

#

cp_info_header =
(off = 0)	U_SHORT		# reserved
(off = 2)	num_fonts
(off = 4)	bitmap_size	# size of font data (all <num_fonts> sizes)
 (size = 6)

num_fonts	= U_SHORT
 (size = 2)
bitmap_size	= U_SHORT		# eg. for width=8 :  16->4102  08,14,16->9746 (!?)
 (size = 2)

cp_fontdata	= {raw_fontdata}	# font data for all <num_fonts> sizes

#

raw_fontdata =	{char_data}*256

char_data =	{BYTE}*<fontheight>