PROBLEM:
	Cannot determine user name by reference to /etc/passwd. Failure
	to determine the user name causes some programs (notably programs
	built with Informix 4GL) to exit.

REASON:
	The functions that parse /etc/passwd are sensitive under SCO
	at least. A corrupt line is taken to be the end of the file
	and further entries are ignored. Any line which has a null
	home directory counts as a corrupt line for SCO.

SOLUTION:
	Ensure that all lines in /etc/passwd have home directories
	specified. Simply edit /etc/passwd with an editor.


# -----------------------------------------------------------------------

PROBLEM:
	Cannot connect to X server using local connection.

REASON:
	Some versions of X libraries seem to require 'localhost' to be
	allowed to connect to the X server even if we aren't using a
	TCP/IP connection. The X libraries used for VSI*FAX require
	this, the X libraries used for WordPerfect don't.

SOLUTION:
	Explicitly allow 'localhost' to connect to the X server using
	'xhost localhost'.


# -----------------------------------------------------------------------

PROBLEM:
	Installation disks claim to be Unix tar format but they appear
	to be blank MS-DOS format disks!

REASON:
	Unix provides two floppy devices, the normal floppy devices
	that we all know and love (and which are listed as the Xenix
	compatible devices in the man page) and a second set which
	skips the first track of the disk. For some reason a few vendors
	seem to use this second set when preparing distribution disks.
	WordPerfect seem to do this. Linux currently only supports
	the normal floppy devices.

SOLUTION:
	You have to skip the first track by hand and read each disk
	individually. Try 'dd if=/dev/fd0 bs=18k skip=1 | tar xfv -'
	for a 3.5" high density disk. Change the 18k to 15k for a
	5.25" high density disk.

# -----------------------------------------------------------------------

PROBLEM:
	A script bombs out with an unexpected EOF looking for "'". This
	only happens on Linux.

REASON:
	There is a bug in the script which is only a problem on Linux.
	Take the following example:

		count=`ls | wc | awk '{ printf "%05d", $1 }`

	Note the missing "'" at the end of the awk statment. The /bin/sh
	supplied with SCO will assume (in this case correctly) that the
	"'" should have occurred immediately before the closing "`" and
	the expression will succeed. The /bin/sh used with Linux (normally
	bash) does not make this assumption and gives an error message.

SOLUTION:
	Either fix the script or arrange to have it run by a SYSV shell
	rather than bash.

# -----------------------------------------------------------------------

PROBLEM:
	Test complains that a numeric argument is required before -eq, -le
	etc.

REASON:
	The GNU shellutils test and the test built in to bash which are
	the versions of test used under Linux do not accept a null
	argument as equivalent to 0 so 'test "" -le 5' will give an error.
	Under SCO a null argument is taken as equivalent to 0 so the
	statement would be evaluated as 'test 0 -le 5'.

SOLUTION:
	Fix the script to check arguments for validity before using them
	or provide a fixed version of test and ensure that it is used in
	preference to any shell built in.

# -----------------------------------------------------------------------

PROBLEM:
	Some X fonts supplied with packages appear corrupt.

REASON:
	These are probably snf fonts. The XFree86 X server used with
	Linux appears to fail to load some snf fonts silently and displays
	garbage. Pcf fonts work ok and should be used where possible.

SOLUTION:
	If you only have the snf fonts all you can do is to try asking the
	vendor for pcf versions or the bdf definitions.

	If you have the bdf definitions (WordPerfect ship them) then you
	can build a pcf set using:

		# for bdffn in *.bdf
		> do
		>   fn=`basename $bdffn .bdf`
		>   [ -r $fn.snf ] && mv $fn.snf $fn.snf.old
		>   bdftopcf -o $fn.pcf $bdffn
		> done
		# mkfontdir .
		# xset fp rehash

# -----------------------------------------------------------------------

PROBLEM:
	Function keys don't work.

REASON:
	The default sequences for function keys under Linux are "sort of"
	VT10x like whereas you have probably told your application that
	you are running on an ansi or SCO console.

SOLUTION:
	It is probably best to run as an ansi screen - especially if
	colour is used - so use the "loadkeys" program to load the
	SCO.map keyboard description from the PROD.Patches directory.

# -----------------------------------------------------------------------

PROBLEM:
	Line drawing characters come out as weird symbols.

REASON:
	The application believes you are using a console screen with a
	PC character set and is using 8-bit codes rather than escape
	sequences. Linux defaults to an ISO8859-1 character set though.

SOLUTION:
	Linux can be switched to PC character set mode with the escape
	sequence ESC-(-U. Arrange to have this sequence sent either
	before the application is started or as part of the initialisation
	that the application does. You can restore the ISO character
	set afterwards with ESC-(-K.
