Copyright 1997-2009,2010 by Thomas E. Dickey

Contents


What is XTERM?

From the manual page:
The xterm program is a terminal emulator for the X Window System. It provides DEC VT102 and Tektronix 4014 compatible terminals for programs that can't use the window system directly. If the underlying operating system supports terminal resizing capabilities (for example, the SIGWINCH signal in systems derived from 4.3bsd), xterm will use the facilities to notify programs running in the window whenever it is resized.

Who wrote XTERM?

I've been working on xterm since early 1996 (see my changelog for details).

But the program is much older than that.

A lot of people, cited at the bottom of the manual page wrote the original xterm program, maintained by the X Consortium (later part of the Open Group – I'm well aware of the distinction, but am citing when the work was done, not who the current owner may be). There is no changelog, and it is not clear who did what. Email from Jim Gettys provides some background:

Cast of thousands...

To give a bit of history, xterm predates X!

It was originally written as a stand-alone terminal emulator for the VS100 by Mark Vandevoorde, as my coop student the summer that X started.

Part way through the summer, it became clear that X was more useful than trying to do a stand alone program, so I had him retarget it to X. Part of why xterm's internals are so horrifying is that it was originally intended that a single process be able to drive multiple VS100 displays. Don't hold this against Mark; it isn't his fault.

I then did a lot of hacking on it, and merged several improved versions from others back in.

Notable improvements include the proper ANSI parser, that Bob McNamara did.

The Tek 4010 support came from a guy at Smithsonian Astrophysical Observatory whose name slips my mind at the moment.

Ported to X11 by Loretta Guarino.

Then hacked on at the X Consortium by uncounted people.

Email from Doug Mink provides more background:
I was checking out the newly revised AltaVista search engine to see what was on the net about xterm, and I found your pages. I can add to the FAQ in that I was the "guy at the Smithsonian Astrophysical Observatory" Jim Gettys refers to. I am listed at the end of the man page under authors. What happened was that I was hired by SAO (after leaving the research staff at MIT) in October 1985 to write analysis software for the Spacelab 2 Infrared Telescope which was to fly on the Space Shuttle in 1985 less than six months after I was hired. I came with a tar tape full of software I had written for Unix and Tektronix terminals, but I was presented with a VS100 terminal which had an early version (X6 or so) of xterm, with no graphics capabilities. SAO is at Harvard, across Cambridge from MIT, where Jim Gettys was detailed from DEC to the X project, and Jim had connections with SAO, having worked here after college (MIT, where we had both worked at the observatory at various times); he was still sharing an apartment with an SAO colleague of mine, too. Anyway, everyone decided that since I knew Tektronix commands pretty well, and our group desparately needed the graphics capabilities, it would be a good use of my time to implement a Tektronix terminal emulator under X. So I set to work learning more C--I had only written a couple of wrappers to C I/O routines so I could use them with my Fortran software--and wrote a Tektronix emulator. The only X documentation at the time was the code itself. While I was at it, I wrote an improved Tektronix emulator for our Imagen laser printer which used the full resolution of that 300 dpi printer instead of the effective 100 dpi (i.e. jaggy) emultator distributed with the printer. The original xterm Tek emulator shared a window with the VT100 emulator, much like on the VT240 terminals which I had been using at MIT before I came to Harvard. With a VAX 750 running several VS100's, window creation was sloowww, so sharing a window was the quickest way to do things, and all of my software was written for that mode of operation, anyway. While I wrote the emulator so that my software would work on it, it was tested by the X group against a BBN graphics package, the name of which slips my mind right now.

Anyway, 15 years later, I am still using xterm and some of the same mapping software I wrote the emulator for. And I am still at the Smithsonian Astrophysical Observatory.

This FAQ is oriented toward the version of xterm originally distributed with XFree86 (also known as "new xterm", with a corresponding terminal description "xterm-new"), which was based on the X11R6.3 xterm, with the addition of ANSI color and VT220 controls.

What platforms does it run on?

Xterm runs in all of the implementations of X11. I've built and run these since I started working on xterm:

The older configurations have X11R5 libraries. Only minor changes are needed to make xterm work on those systems. However, with X11R6 you can obtain better locale support, as well as new features such as the active icon. X11R7... not much to say there.

What is the latest version?

The most recent (and well supported) version of xterm is the one that I maintain:

What versions are available?

There are several other versions of xterm, based on xterm's source. These include There are similar programs not based on xterm's source, which are compatible to different degrees. These include Some of these use the VTE widget. Since that supplies most of the terminal emulation, the remaining differences between programs using VTE tend to be at the level of the window manager (menus, borders, etc.). Other (older) programs which are based on reusable widgets include dtterm and emu.

(I am aware of a few others, such as xcterm, but have not seen a working version of these).

There were for some time two versions of XFree86 xterm. Starting with my patch 88, there were the stable (beta) and unstable (alpha) versions, which currently are XFree86 3.3.6 and XFree86 4.x, respectively. I made only critical changes to the stable version since patch 88; ongoing development (including all non-critical fixes) was focused on the "unstable".

XFree86 4.0 should have incorporated my patch 131, but it was overlooked at the last moment (though it was listed in the ChangeLog, the patch itself was not applied). Unfortunately, the patch 130 version which was released renders colors incorrectly on most platforms, in particular FreeBSD. Ironically, the change in patch 129 which introduced this bug was designed to work around a bug which I have seen only running with FreeBSD. To compound the irony, there was some resistance at the time (2000/4/5) by that organization to incorporating the fix because it might be confused with the 4.0 release version.

How do I ...

Not really problems, but frequently asked questions (the point of this, after all):

How do I change the font size?

xterm uses fonts given as resource settings. You can switch between these fonts at runtime, using a menu. This is documented in the manpage, in the MENUS section.

X Consortium xterm provides popup menus, by pressing the control key together with the mouse button. Control right mouse button pops up the VT FONTS menu, from which you can select fonts that are specified in xterm's resources. Usually these are in increasing order of size.

XFree86 xterm provides the menu, plus a feature adapted from rxvt: pressing the shifted keypad plus or minus keys steps through the font menu selections, in order of their size.

xterm's manpage does not document the syntax for X resources; it is done in the X documentation. If you are instead asking about a problem displaying a given font, it may be due to a problem with your resource settings.

How do I print the screen?

That depends on why you want to print it.

If you want a trace of an interactive session, you should use the script program. It records every character sent to the screen, recording them in a file typescript. There are two drawbacks to this approach:

Well, what about logging? Some versions of xterm support logging to a file. In fact XFree86 xterm does. Logging was dropped from X Consortium xterm during X11R5 due to security concerns. Those were addressed, but logging was not reinstated (in fact there is a related bug in xterm). Some people prefer this, because it is convenient: you can start and stop logging a popup menu entry. However Both script and logging are useful for recording, but they require interpretation to make sense of the trace. You probably would not send that trace to a printer (not twice, anyway).

If you want to print the contents of the screen, XFree86 xterm implements, as part of the VT100 emulation, an "attached" printer.

There are limitations and tradeoffs using the "attached" printer, because it is an emulation: If you use the popup menu to print the screen, this will close the printer pipe unless it was already opened by the application running in xterm.

How do I set up function keys?

With XFree86 xterm, this is relatively simple. So I'll answer that first.

With X Consortium xterm, you have partial support for DEC VTxxx function keys. Function keys F1 to F12 correspond to DEC's F1 to F12 (sort of). Actually, DEC's VT220 terminals do not have codes for F1 through F5. They are reserved for local functions. And the VT220 (and up) terminals have 20 function keys. So you cannot do anything with the F13 through F20 (i.e., DO, HELP and SELECT). Finally, though xterm is reputed to be VT100-compatible, it has no support for the VT100 keypad (PF1 to PF4, and the "," key).

XFree86 xterm changes the X Consortium codes for F1 to F4 to match the VT100 PF1 to PF4, except when the emulation level is VT220 and up. In this case, it generates the same F1 to F4 codes as X Consortium xterm. Moreover, it adds a new resource sunKeyboard, which tells the program whether it has only 12 function keys (i.e., a Sun or PC keyboard). If so (this is selectable from the popup menu), you can use the control key with F1 to F12 to get F13 to F24, and use the "+" key on the keypad as an alias for "," (comma).

The emulation level for XFree86 xterm is set via the resource decTerminalID, e.g., to 220 for a VT220. Once set, applications can set the emulation level up or down within that limit. DEC's terminals are configured in much the same way by a setup option.

That is the simple way, using a couple of new resources. The traditional way to get function keys involves translations. I have seen a few postings on the newsgroups that do this. Here is one from Bruce Momjian <root@candle.pha.pa.us> for a VT220:


	xterm $XTERMFLAGS +rw +sb +ls $@ -tm 'erase ^? intr ^c' \
		-name vt220 -title vt220 -tn xterm-220 "$@" &

with the corresponding resources:


	XTerm*VT100.translations: #override \n\
		<Key>Home: string(0x1b) string("[3~") \n \
		<Key>End: string(0x1b) string("[4~") \n
	vt220*VT100.translations: #override \n\
	~Shift	<Key>F1: string(0x1b) string("OP") \n \
	~Shift	<Key>F2: string(0x1b) string("OQ") \n \
	~Shift	<Key>F3: string(0x1b) string("OR") \n \
	~Shift	<Key>F4: string(0x1b) string("OS") \n \
	~Shift	<Key>F5: string(0x1b) string("[16~") \n \
	~Shift	<Key>F6: string(0x1b) string("[17~") \n \
	~Shift	<Key>F7: string(0x1b) string("[18~") \n \
	~Shift	<Key>F8: string(0x1b) string("[19~") \n \
	~Shift	<Key>F9: string(0x1b) string("[20~") \n \
	~Shift	<Key>F10: string(0x1b) string("[21~") \n \
	~Shift	<Key>F11: string(0x1b) string("[28~") \n \
	~Shift	<Key>F12: string(0x1b) string("[29~") \n \
	Shift	<Key>F1: string(0x1b) string("[23~") \n \
	Shift	<Key>F2: string(0x1b) string("[24~") \n \
	Shift	<Key>F3: string(0x1b) string("[25~") \n \
	Shift	<Key>F4: string(0x1b) string("[26~") \n \
	Shift	<Key>F5: string(0x1b) string("[K~") \n \
	Shift	<Key>F6: string(0x1b) string("[31~") \n \
	Shift	<Key>F7: string(0x1b) string("[31~") \n \
	Shift	<Key>F8: string(0x1b) string("[32~") \n \
	Shift	<Key>F9: string(0x1b) string("[33~") \n \
	Shift	<Key>F10: string(0x1b) string("[34~") \n \
	Shift	<Key>F11: string(0x1b) string("[28~") \n \
	Shift	<Key>F12: string(0x1b) string("[29~") \n \
		<Key>Print: string(0x1b) string("[32~") \n\
		<Key>Cancel: string(0x1b) string("[33~") \n\
		<Key>Pause: string(0x1b) string("[34~") \n\
		<Key>Insert: string(0x1b) string("[2~") \n\
		<Key>Delete: string(0x1b) string("[3~") \n\
		<Key>Home: string(0x1b) string("[1~") \n\
		<Key>End: string(0x1b) string("[4~") \n\
		<Key>Prior: string(0x1b) string("[5~") \n\
		<Key>Next: string(0x1b) string("[6~") \n\
		<Key>BackSpace: string(0x7f) \n\
		<Key>Num_Lock: string(0x1b) string("OP") \n\
		<Key>KP_Divide: string(0x1b) string("Ol") \n\
		<Key>KP_Multiply: string(0x1b) string("Om") \n\
		<Key>KP_Subtract: string(0x1b) string("OS") \n\
		<Key>KP_Add: string(0x1b) string("OM") \n\
		<Key>KP_Enter: string(0x1b) string("OM") \n\
		<Key>KP_Decimal: string(0x1b) string("On") \n\
		<Key>KP_0: string(0x1b) string("Op") \n\
		<Key>KP_1: string(0x1b) string("Oq") \n\
		<Key>KP_2: string(0x1b) string("Or") \n\
		<Key>KP_3: string(0x1b) string("Os") \n\
		<Key>KP_4: string(0x1b) string("Ot") \n\
		<Key>KP_5: string(0x1b) string("Ou") \n\
		<Key>KP_6: string(0x1b) string("Ov") \n\
		<Key>KP_7: string(0x1b) string("Ow") \n\
		<Key>KP_8: string(0x1b) string("Ox") \n\
		<Key>KP_9: string(0x1b) string("Oy") \n

	!	<Key>Up: string(0x1b) string("[A") \n\
	!	<Key>Down: string(0x1b) string("[B") \n\
	!	<Key>Right: string(0x1b) string("[C") \n\
	!	<Key>Left: string(0x1b) string("[D") \n\

	*visualBell:	true
	*saveLines:    1000
	*cursesemul:	true
	*scrollKey: true
	*scrollBar: true
Note that real VT220 terminals use shifted function keys to mean something different: the user-programmable keys (i.e., DECUDK). XFree86 xterm supports this, but the translations do not (they're using shift to select F13 to F20).

Here's another one, from Robert Ess <ress@spd.dsccc.com>:


	#!/bin/sh

	#		vax
	# 09-17-96	Bob Ess      - initial creation
	# 09-26-96	Shig Katada  - Additional keybindings
	#
	#		Script file to incorporate keybindings and command line
	#		options for connecting to a VAX node

	# Usage statement
	Usage(){
		echo
		echo " Usage  : vax -options"
		echo
		echo " Options: -80   for 80 column terminal"
		echo "          -132  for 132 column terminal"
		echo "          -fg colorname"
		echo "          -bg colorname"
		echo "          -fn fontname"
		echo "          -fb bold fontname"
		echo "          -host [altair] [devel] [leonis] [castor]"
		echo ""
		echo " Example: \"vax -80 -fg white -bg black -fn 9x15 -fb 9x15b -host castor\""
		echo "          Starts a VAX session with an 80 column terminal"
		echo "          with a black background, white foreground, a normal"
		echo "          font of 9x15 and a bold font of 9x15b, and connects"
		echo "          to the node 'castor'"
		echo
		echo "          If you need additional help, please call Workstation"
		echo "          Services at x92396."
		echo
		exit 1
	}


	# Default to a black foreground with a white background.
	# Use the 9x15 and 9x15bold fonts. Connect to castor by default.
	#
	FG=black
	BG=white
	HOST=castor
	FONT=9x15
	BFONT=9x15bold
	COLS=80


	# Parse the command line arguments
	#
	while [ $# != 0 ];
	do
		case $1 in
			-80)	COLS=80
				FONT=spc12x24c
				BFONT=spc12x24b
				shift
				;;
			-132)	COLS=132
				FONT=9x15
				BFONT=9x15b
				shift
				;;
			-fg)	shift
				FG=$1
				shift;;
			-bg)	shift
				BG=$1
				shift;;
			-fn)	shift
				FONT=$1
				shift;;
			-fb)	shift
				BFONT=$1
				shift;;
			-host)	shift
				HOST=$1
				shift;;
			-help)	Usage;;
			*)	Usage;;
		esac
	done


	xterm  -title "VAX" -sb -sl 1200 -geo ${COLS}x24 -fg ${FG} -bg ${BG} \
		-cr red -fn ${FONT} -fb ${BFONT} -xrm \
		'XTerm*VT100.translations:     #override \n\
		<Key>Insert:            string(\001) \n\
		Shift <Key>Up:          scroll-back(1,lines) \n\
		Shift <Key>Down:        scroll-forw(1,lines) \n\
		Shift <Key>Right:       string(0x1b) string("f") \n\
		Shift <Key>Left:        string(0x1b) string("b") \n\
		Shift <Key>Delete:      string(0x1b) string(0x08) \n\
		Shift <Key>Tab:         string(0x1b) string("*") \n\
		<Key>0x1000FF0D:        scroll-back(1,page) \n\
		<Key>0x1000FF0E:        scroll-forw(1,page) \n\
		<Key>0x1000FF09:        string(\010) \n\
		<Key>0x1000FF0A:        string(\005) \n\
		<Key>BackSpace:         string(0xff) \n\
		<Key>Select:            select-start() \n\
		<Key>0x1000FF02:        select-end(PRIMARY,CUT_BUFFER0) \n\
		Meta <Key>0x1000FF02:   select-end(CLIPBOARD) \n\
		<Key>0x1000FF04:        insert-selection(PRIMARY,CUT_BUFFER0) \n\
		Meta <Key>0x1000FF04:   insert-selection(CLIPBOARD) \n\
		<Key>F1:                string(0x1b) string("OP") \n\
		<Key>F2:                string(0x1b) string("OQ") \n\
		<Key>F3:                string(0x1b) string("OR") \n\
		<Key>F4:                string(0x1b) string("OS") \n\
		<Key>F5:                string(0x1b) string("OA") \n\
		<Key>F11:               string(0x1b) string("[23~") \n\
		<Key>F12:               string(0x1b) string("[24~") \n\
		<Key>KP_0:              string(0x1b) string("Op") \n\
		<Key>KP_1:              string(0x1b) string("Oq") \n\
		<Key>KP_2:              string(0x1b) string("Or") \n\
		<Key>KP_3:              string(0x1b) string("Os") \n\
		<Key>KP_4:              string(0x1b) string("Ot") \n\
		<Key>KP_5:              string(0x1b) string("Ou") \n\
		<Key>KP_Divide:         string(0x1b) string("OP") \n\
		<Key>KP_Multiply:       string(0x1b) string("[29~") \n\
		<Key>KP_Enter:          string(0x1b) string("OM") \n\
		<Key>KP_Subtract:       string(0x1b) string("Om") \n\
		<Key>KP_Add:            string(0x1b) string("Ol") \n\
		<Key>KP_Decimal:        string(0x1b) string("On") \n\
		<Btn1Down>:             select-start() \n\
		<Btn1Motion>:           select-extend() \n\
		<Btn1Up>:               select-end(PRIMARY,CUT_BUFFER0) \n\
		Button1<Btn2Down>:      select-end(CLIPBOARD) \n\
		Button1<Btn2Up>:        ignore()' \
		-e telnet $HOST &
Finally (for the moment) is a further modification of Robert Ess's script by Erik Ahlefeldt, <oahlefel@metz.une.edu.au>. From his readme file, for vmsterm:
This script is for people who wish to connect from a Linux or Unix computer to a VMS computer using telnet and get a good VT100 or VT220 emulation. The key mappings have been specifically designed to emulate the VT terminal auxiliary numeric keypad, so that you can use VMS EDT and TPU editors, as well as the many VMS applications use keys PF1 to PF4. The script should work with any recent version of Xterm using a standard extended IBM PC keyboard or a Sun keyboard.

About the keymappings. First the auxiliary numeric keypad. My prime objective with these mappings was to produce a setup that I could use with the EDT and TPU editors which make extensive use of the numeric keypad. The top row of keys PC numeric keypad (Num Lock, Divide, Multiply, Subtract) are where you find PF1, PF2, PF3, PF4 on a VT keyboard, so I have mapped them to PF1 thru PF4. The PC numeric keypad Add key (+) takes up the space of two keys which are Minus and Comma on the VT keyboard – I have mapped it to Comma (Delete Character in the EDT editor). I have then used the PC Pause key to map to VT key Minus (Delete Word in the EDT editor). The remaining keys on the auxiliary numeric keypad are the same for PC and VT.

The six keys between the main and numeric keypads on the PC (Insert, Home, Page Up, Delete End, Page Down) are usually mapped to the VT keys by either position or by (approximate) function. As I rarely use these keys I have mapped them by function as follows: PC key Insert to VT Insert Here, PC Home to VT Find, PC Page Up to VT Prev, PC Delete to VT Remove, PC End to VT Select, PC Page Down to VT Next.

Function keys.
There are 12 function keys on the PC keyboard and 20 on the VT keyboard, so I map PC F1 thru F12 to VT F1 thru F12 (except for F1 thru F5 as noted below) and PC Shift F1 thru Shift F10 to VT F11 thru F20.

The VT keys F1 thru F5 are local hardware function keys so there is nothing to emulate, however some PC to VT emulations in the past have mapped PF1 thru PF4 here, so I have done that too, even though they are already mapped on the auxiliary numeric keypad.

Xterm functionality.
You lose some xterm functions when you remap the keyboard, however this script implements a scroll back buffer of 1000 lines which you scroll through using Shift and Up (a.k.a. Up Arrow or Cursor Up key) or Shift and Down.
a summary of the keyboard mapping:
		PC Key     maps to   VT Key.
		------               ------
		F1                   PF1
		F2                   PF2
		F3                   PF3
		F4                   PF4
		F5                   unused
		F6                   F6
		F7                   F7
		F8                   F8
		F9                   F9
		F10                  F10
		F11                  F11
		F12                  F12
		Shift F1             F11
		Shift F2             F12
		Shift F3             F13
		Shift F4             F14
		Shift F5             F15 (Help)
		Shift F6             F16 (Do)
		Shift F7             F17
		Shift F8             F18
		Shift F9             F19
		Shift F10            F20
		Shift F11            F11
		Shift F12            F12
		Print                Help (F15)
		Cancel               Do   (F16)
		Pause                Keypad Minus

		Insert               Insert Here
		Delete               Remove
		Home                 Find
		End                  Select
		Prior                Prev
		Next                 Next
		BackSpace            BackSpace (sends DEL - ascii 127)

		Num_Lock             PF1
		KP_Divide            PF2
		KP_Multiply          PF3
		KP_Subtract          PF4
		KP_Add               Keypad Comma
		KP_Enter             Enter
		KP_Decimal           Period
		KP_0                 Keypad 0
		KP_1                 Keypad 1
		KP_2                 Keypad 2
		KP_3                 Keypad 3
		KP_4                 Keypad 4
		KP_5                 Keypad 5
		KP_6                 Keypad 6
		KP_7                 Keypad 7
		KP_8                 Keypad 8
		KP_9                 Keypad 9
		Up                   Up
		Shift Up             Scroll Back
		Down                 Down
		Shift Down           Scroll Forward
		Right                Right
		Left                 Left
and the script:

	#!/bin/sh
	#               vmsterm
	#               from an original script by Bob Ess
	#               key translations by Erik Ahlefeldt
	#
	#               Script file using Xterm and telnet to connect to a VMS host
	#               and give a decent vt220 emulation.
	#
	# Usage statement
	Usage(){
		echo
		echo " Usage  : vmsterm -options"
		echo
		echo " Options: -80   for 80 column terminal"
		echo "          -132  for 132 column terminal"
		echo "          -bg colorname"
		echo "          -fg colorname"
		echo "          -fn fontname"
		echo "          -fb bold fontname"
		echo "          -host [crusher.saltmine.com] [earth] [192.168.7.7]"
		echo ""
		echo " Example: \"vmsterm -80 -fg white -bg black -fn 9x15 -fb 9x15b -host earth\""
		echo "          Starts a VMS session with an 80 column terminal"
		echo "          with a black background, white foreground, a normal"
		echo "          font of 9x15 and a bold font of 9x15b, and connects"
		echo "          to the node 'earth'"
		echo ""
		echo " Example: \"vmsterm -host earth\""
		echo "          Starts a VMS session with default terminal settings "
		echo ""
		echo " Example: \"vmsterm -help\""
		echo "          Displays vmsterm options "
		echo
		exit 1
	}

	# Default to a black foreground with a white background.
	# Use the 9x15 and 9x15bold fonts. Connect to 192.168.3.3 by default.
	#
	FG=black
	BG=white
	HOST=192.168.3.3
	FONT=9x15
	BFONT=9x15bold
	COLS=80

	# Parse the command line arguments
	#
	while [ $# != 0 ];
	do
		case $1 in
			-80)    COLS=80
				FONT=spc12x24c
				BFONT=spc12x24b
				shift
				;;
			-132)   COLS=132
				FONT=9x15
				BFONT=9x15b
				shift
				;;
			-fg)    shift
				FG=$1
				shift;;
			-bg)    shift
				BG=$1
				shift;;
			-fn)    shift
				FONT=$1
				shift;;
			-fb)    shift
				BFONT=$1
				shift;;
			-host)  shift
				HOST=$1
				shift;;
			-help)  Usage;;
			*)      Usage;;
		esac
	done

	xterm	-title "VMSTERM" -sb -sl 1000 -geo ${COLS}x24 -fg ${FG} -bg ${BG} \
		-cr blue -fn ${FONT} -fb ${BFONT} -xrm \
		'XTerm*VT100.translations: #override \n \
		~Shift	<Key>F1:	string(0x1b)	string("OP") \n \
		~Shift	<Key>F2:	string(0x1b)	string("OQ") \n \
		~Shift	<Key>F3:	string(0x1b)	string("OR") \n \
		~Shift	<Key>F4:	string(0x1b)	string("OS") \n \
		~Shift	<Key>F5:	string("Break") \n \
		~Shift	<Key>F6:	string(0x1b)	string("[17~") \n \
		~Shift	<Key>F7:	string(0x1b)	string("[18~") \n \
		~Shift	<Key>F8:	string(0x1b)	string("[19~") \n \
		~Shift	<Key>F9:	string(0x1b)	string("[20~") \n \
		~Shift	<Key>F10:	string(0x1b)	string("[21~") \n \
		~Shift	<Key>F11:	string(0x1b)	string("[23~") \n \
		~Shift	<Key>F12:	string(0x1b)	string("[24~") \n \
		Shift   <Key>F1:	string(0x1b)	string("[23~") \n \
		Shift   <Key>F2:	string(0x1b)	string("[24~") \n \
		Shift   <Key>F3:	string(0x1b)	string("[25~") \n \
		Shift   <Key>F4:	string(0x1b)	string("[26~") \n \
		Shift   <Key>F5:	string(0x1b)	string("[28~") \n \
		Shift   <Key>F6:	string(0x1b)	string("[29~") \n \
		Shift   <Key>F7:	string(0x1b)	string("[31~") \n \
		Shift   <Key>F8:	string(0x1b)	string("[32~") \n \
		Shift   <Key>F9:	string(0x1b)	string("[33~") \n \
		Shift   <Key>F10:	string(0x1b)	string("[34~") \n \
		Shift   <Key>F11:	string(0x1b)	string("[28~") \n \
		Shift   <Key>F12:	string(0x1b)	string("[29~") \n \
			<Key>Print:	string(0x1b)	string("[28~") \n \
			<Key>Cancel:	string(0x1b)	string("[29~") \n \
			<Key>Pause:	string(0x1b)	string("Om") \n \
			<Key>Insert:	string(0x1b)	string("[2~") \n \
			<Key>Delete:	string(0x1b)	string("[3~") \n \
			<Key>Home:	string(0x1b)	string("[1~") \n \
			<Key>End:		string(0x1b)	string("[4~") \n \
			<Key>Prior:	string(0x1b)	string("[5~") \n \
			<Key>Next:	string(0x1b)	string("[6~") \n \
			<Key>BackSpace:	string(0x7f)	\n \
			<Key>Num_Lock:	string(0x1b)	string("OP") \n \
			<Key>KP_Divide:	string(0x1b)	string("OQ") \n \
			<Key>KP_Multiply: string(0x1b)	string("OR") \n \
			<Key>KP_Subtract: string(0x1b)	string("OS") \n \
			<Key>KP_Add:	string(0x1b)	string("Ol") \n \
			<Key>KP_Enter:	string(0x1b)	string("OM") \n \
			<Key>KP_Decimal: string(0x1b)	string("On") \n \
			<Key>KP_0:	string(0x1b)	string("Op") \n \
			<Key>KP_1:	string(0x1b)	string("Oq") \n \
			<Key>KP_2:	string(0x1b)	string("Or") \n \
			<Key>KP_3:	string(0x1b)	string("Os") \n \
			<Key>KP_4:	string(0x1b)	string("Ot") \n \
			<Key>KP_5:	string(0x1b)	string("Ou") \n \
			<Key>KP_6:	string(0x1b)	string("Ov") \n \
			<Key>KP_7:	string(0x1b)	string("Ow") \n \
			<Key>KP_8:	string(0x1b)	string("Ox") \n \
			<Key>KP_9:	string(0x1b)	string("Oy") \n \
		~Shift	<Key>Up:		string(0x1b)	string("[A") \n \
		Shift	<Key>Up:		scroll-back(1,lines) \n \
		~Shift	<Key>Down:	string(0x1b)	string("[B") \n \
		Shift	<Key>Down:	scroll-forw(1,lines) \n \
			<Key>Right:	string(0x1b)	string("[C") \n \
			<Key>Left:	string(0x1b)	string("[D")' \
		-e telnet $HOST

How do I set the title?

The control sequences for doing this are documented in ctlseqs.ms.

The usual context for this question is setting the title according to the current working directory. People post answers to this periodically on the newsgroups. Here is one that I have seen, from Roy Wright <nobody@roystoy.dseg.ti.com>. In your /etc/profile after:


	if [ "$SHELL" = "/bin/pdksh" -o "$SHELL" = "/bin/ksh" ]; then
		PS1="! $ "
	elif [ "$SHELL" = "/bin/zsh" ]; then
		PS1="%m:%~%# "
	elif [ "$SHELL" = "/bin/ash" ]; then
		PS1="$ "
	else
		PS1='\u@\h:\w\$ '
	fi

add:


	if [ "$TERM" = "xterm" ]; then
		PS1="\033]2;\u@\h:\w\007bash$ "
	fi

The terminator "\007" is a problem area. Xterm historically uses this character, though it is non-ANSI. The "correct" character should be a "\233" string terminator, or "\033\\", which is the 7-bit equivalent. XFree86 xterm recognizes either (the "\007" or string terminator); waiting for the first of these.

You may have resource or environment problems that prevent you from setting the title at all. Newer xterms (starting somewhere in X11R5) use the $LANG variable. If your locale is incorrectly installed, you will be unable to set the xterm's title. As noted by Mikhail Teterin <mi@rtfm.ziplink.net>: Make sure that the locale (LANG and/or LOCALE environment variable) is known to X Window System. Check ${X11ROOT}/lib/X11/locale.* for it. If it is not listed in either one of the files, find the nearest match and add an alias to it. Restart X if you have made changes.

On a related note, some people want to know how to read the title from an xterm. This works for XFree86 xterm and dtterm, but not for other variations:


	#!/bin/ksh
	# Echo the current X term title bar to standard output.
	# Written by Icarus Sparry <icarus@bath.ac.uk> 11 Apr 1997
	#
	exec </dev/tty
	old=$(stty -g)
	stty raw -echo min 0  time ${1-10}
	print "\033[21t\c" > /dev/tty
	IFS='' read -r a
	stty $old
	b=${a#???}
	print -R "${b%??}"
But it is possible to avoid escape sequences altogether (from Hemant Shah <shah@typhoon.xnet.com>):

	$ xprop -id $WINDOWID | grep WM_NAME
	WM_NAME(STRING) = "this is my title"
	current_title=$(xprop -id $WINDOWID | grep WM_NAME | cut -d= -f2)
Here's another source of information: Xterm-Title HowTo

How do I make the cursor blink?

Standard xterm does not implement a blinking cursor. Some of the variations do: dtterm, GNOME Terminal, and XFree86 xterm (from mid 1999, patch 107).

Frequent problems

My terminal doesn't show box characters

Xterm displays the 7-bit ASCII and VT100 graphic characters (including box corners) using specially arranged fixed-pitch fonts. The first 32 glyph positions (which would correspond to nonprinting control characters) are used to hold the VT100 graphic characters. Some fonts that otherwise look fine (such as courier) do not have glyphs defined for these positions. So they display as blanks. Use xfd to display the font.

XFree86 xterm can form its own line-drawing characters (see patch 90, for example). It does not draw all of the graphic characters, only those that may be done with straight lines. But those are the most used, making most of the fixed-pitch fonts useful for xterm.

You may also have a problem with the terminfo description. As distributed, the X11R6 terminfo for xterm does not have the acsc string defined, so most implementations of curses do not try to use the alternate character set.

Finally, some people confuse the VT100 graphic characters with the VT220 support for DEC technical character set. These are distinct (7-bit) character sets. Xterm currently does not support this.

The bold font is ugly

Xterm lets you directly specify one bold font, which is assumed to correspond to the default font. Older versions of xterm make a fake bold font for the other choices via the fonts menu by drawing the characters offset by one pixel. I modified xterm to ask the font server for a bold font that corresponds to each font (other than the default one). Usually that works well. However, sometimes the font server gives a poor match. Xterm checks for differences in the alignment and size, but the font server may give incorrect information about the font size. The scaled bitmap font feature gives poor results for the smaller fonts. In your X server configuration file, that can be fixed by disabling the feature, e.g., by appending ":unscaled" to the path:

	FontPath	"/usr/lib/X11/fonts/100dpi/:unscaled"
	FontPath	"/usr/lib/X11/fonts/75dpi/:unscaled"
	FontPath	"/usr/lib/X11/fonts/misc/:unscaled"

I see little dots on the screen

Well, I do. Perhaps you do not. It depends on the fonts you choose, and how you use them.

Standard xterm has a "normal" font for which a bold font can be chosen, and several alternative fonts, useful for changing the font size. The alternative fonts do not have corresponding bold fonts. Xterm simulates bold fonts in this case by overstriking the character one pixel offset. That can make an bold character extend into the area that another character occupies. When erasing a bold character from the screen, xterm does not erase the extra pixel. This is corrected in XFree86 xterm, subject to the available fonts (from late 1998, patch 85). For each font, it asks the font server for a corresponding bold font. Your font server may not have the bold font (or it may incorrectly report that it does). But it usually works.

My terminal doesn't display Cyrillic characters

Cyrillic encodings typically use characters in the range 128-159. For a VT220 (or any terminal that follows ISO 6429), those are treated as control characters. Still, some people want to use KOI8-R, etc. I modified xterm in patch 175 to add an option (-k8) and corresponding resource settings to allow them to customize their environment. Here is a sample script and resource file which I use for testing this configuration.

I see boxes instead of characters in uxterm

xterm may show boxes instead of characters if the font that you have selected does not contain those characters. Normally you can fix most of that using the UTF-8 feature, with uxterm. However, your X resource settings may be the source of the problem.

One pitfall to setting X resources is that they allow you to specify wildcards, e.g., the "*" character. When you give a wildcard, the X resource matches any number of levels in the widget hierarchy.

xterm has more than one widget matching "font" at different levels of the hierarchy. There are the popup menus, and there are the fonts used for uxterm. The latter is where an overbroad pattern can cause xterm to use a different font than you expect.

Suppose your resource setting includes this pattern

*VT100*font: fixed
It could be interpreted as this:
*VT100.font: fixed
*VT100.utf8Fonts.font: fixed
xterm uses the utf8Fonts subresources to provide runtime-switchable fonts between IS0-8859-1 (Latin-1) and ISO-10646 (Unicode). Modifying the Unicode font to "fixed" will make most of the characters unavailable (i.e., shown as boxes). If instead your resource looks like
*VT100.font: fixed
it would be unambiguous, and not modify the utf8Fonts value.

The first popup menu is very slow

Some users report that when starting xterm, it is very slow, that their computer's CPU time increases, etc.

This is a longstanding bug in the X libraries. There is a workaround using a resource setting for xterm.

Details

xterm uses the Athena (Xaw) widgets to display popup menus. In the normal case, those are initialized one-by-one as they are first used. If you have configured xterm to use its toolbar configuration, they are all initialized on startup. In the latter, performance problems are more noticeable.

The Athena widgets XawInitializeWidgetSet function goes through several levels down to the X library _XlcAddUtf8LocaleConverters function to call create_tocs_conv and related functions to make a list of character sets from the locale, which is used in menus to get all possible fonts needed for a fontset.

If your current locale uses UTF-8 encoding, this will read a long list of bitmap fonts—everything whose encoding might be useful for displaying the menus. For example, this list (from lcUTF8.c) which dates from around 2000 is the core of the problem:

ISO10646-1, ISO8859-1, ISO8859-2, ISO8859-3, ISO8859-4, ISO8859-5, ISO8859-6, ISO8859-7, ISO8859-8, ISO8859-9, ISO8859-10, ISO8859-11, ISO8859-13, ISO8859-14, ISO8859-15, ISO8859-16, JISX0201.1976-0, TIS620-0, GB2312.1980-0, JISX0208.1983-0, JISX0208.1990-0, JISX0212.1990-0, KSC5601.1987-0, KOI8-R, KOI8-U, KOI8-C, TATAR-CYR, ARMSCII-8, IBM-CP1133, MULELAO-1, VISCII1.1-1, TCVN-5712, GEORGIAN-ACADEMY, GEORGIAN-PS, ISO8859-9E, MICROSOFT-CP1251, MICROSOFT-CP1255, MICROSOFT-CP1256, BIG5-0, BIG5-E0, BIG5-E1, ISO10646-1, ISO10646-1

However, xterm is going to use only the characters shown in the popup menus. It is unlikely that you need Chinese fonts for that.

Solution

xterm's menuLocale resource can be set to an explicit value, e.g., "C" to override the current locale as seen by this initialization debacle.

Limitations

The workaround does not prevent some hacker from "improving" the X libraries still further.

My terminal doesn't recognize color

First, ensure that you have set up xterm to render color. The XFree86 xterm renders color only if you have set resources to do this; the default behavior is monochrome to maintain compatibility with older applications. The manual page describes these resources. I set them in my .Xdefaults file.

Even if you set the resources properly, there may be another application running which prevents xterm from allocating the colors you have specified. But you should see a warning message for this.

Check the terminal description, to see if it is installed properly, e.g., for ncurses, which uses terminfo.

Finally, some applications (that do not interface properly with terminfo or termcap) may need the environment variable $COLORTERM to be set.

Why is my screen size not set?

Well, it may be set, but not correctly. You may notice these symptoms: Xterm knows how big the screen is (of course), and tries to tell your applications (e.g., by invoking ioctl's and sending SIGWINCH). But sometimes it cannot: Most full-screen applications such as vi are designed to use the ioctl calls that return the screen size. When they fail, the applications use the size defined in the terminal's terminfo or termcap description.

You may be able to use the resize program to issue the ioctl's that will notify your application of the actual screen size. This does not always work for the reasons just mentioned. Newer versions of stty let you specify the screen size, though it will not be updated if you resize the xterm window:


	stty rows 24 columns 80
Most full-screen applications also check if the $LINES and $COLUMNS variables are set, using those values to override the terminal description:

	setenv LINES 24
	setenv COLUMNS 80
Why 65 lines? The standard xterm terminfo description specifies 65 lines, perhaps because someone liked it that way. Real VT100's are 24 lines. I once used (and wrote applications for) a Bitgraph terminal, which emulated VT100, but displayed 65 lines.

Why can't I use the pageup/pagedown keys?

Some vendors, e.g,. Sun, have added key translations which make the pageup and pagedown keys talk to the xterm's scrollbar instead of your application. They did the same thing for the home and end keys, thereby obscuring a bug in xterm.

You can override this by specifying your own translations in your resource file. Use the translations in


	/usr/lib/X11/app-defaults/XTerm
as a guide. The relevant section of the app-default file looks like

	*VT100.translations:    #override \
		@Num_Lock<Key>KP_0: string(0)\n\
		@Num_Lock<Key>KP_1: string(1)\n\
		@Num_Lock<Key>KP_2: string(2)\n\
		@Num_Lock<Key>KP_3: string(3)\n\
		@Num_Lock<Key>KP_4: string(4)\n\
		@Num_Lock<Key>KP_5: string(5)\n\
		@Num_Lock<Key>KP_6: string(6)\n\
		@Num_Lock<Key>KP_7: string(7)\n\
		@Num_Lock<Key>KP_8: string(8)\n\
		@Num_Lock<Key>KP_9: string(9)\n\
		@Num_Lock<Key>KP_Add: string(+)\n\
		@Num_Lock<Key>KP_Decimal: string(.)\n\
		@Num_Lock<Key>KP_Divide: string(/)\n\
		@Num_Lock<Key>KP_Enter: string(\015)\n\
		@Num_Lock<Key>KP_Equal: string(=)\n\
		@Num_Lock<Key>KP_Multiply: string(*)\n\
		@Num_Lock<Key>KP_Subtract: string(-)\n\
		<Key>Prior:scroll-back(1,page)\n\
		<Key>Next:scroll-forw(1,page)\n\
		<Key>F16: start-extend() select-end(PRIMARY, CUT_BUFFER0, CLIPBOARD) \n\
		<Key>F18: insert-selection(PRIMARY, CLIPBOARD) \n\
		<Key>F27: scroll-back(100,page) \n\
		<Key>R13: scroll-forw(100,page) \n\
		<Key>Home: scroll-back(100,page) \n\
		<Key>End: scroll-forw(100,page) \n
For example, a more-specific pattern for the resource name lets you override:

	XTerm*VT100.translations:       #override \n\
		~Shift<Key>Home:  string(\033[1~)\n\
		~Shift<Key>End:   string(\033[4~)\n\
		~Shift<Key>Prior: string(\033[5~)\n\
		~Shift<Key>Next:  string(\033[6~)\n\
		Shift<Key>Prior:  scroll-back(1,page) \n\
		Shift<Key>Next:   scroll-forw(1,page) \n\
		Shift<Key>Home:   scroll-back(100,page) \n\
		Shift<Key>End:    scroll-forw(100,page) \n
makes the home/end and pageup/pagedown keys usable by your editor, while leaving their shifted equivalents available for the scrollbar.

Why can't I use the home/end keys?

This is a long story, unless you are referring to X Consortium xterm. That program is simply broken in this respect.

At the beginning, when the home/end keys were fixed for XFree86 xterm (in early 1996), there was some discussion regarding what the escape sequences should be for those keys (for the 6-key editing keypad). Those were chosen as "PC-style" codes (like SCO "ansi"), i.e.,

	ESC [ H
	ESC [ F
for normal mode, and
	ESC O H
	ESC O F
for cursor application mode.

That style of coding fit easily into the existing logic of xterm. It was not my change, and (because xterm should be based upon standards), I did question this, and asked the opinion of the person who was at that time developing rxvt. He had chosen a layout based on DEC's VT220 terminals, though the key labels on the typical PC keyboard did not match. At that point, neither of us knew enough to make a good case for this.

Somewhat later I could see that xterm had a number of undocumented extensions to support the VT220-style (pre-ISO 2022) character sets. I decided to complete the functionality by making xterm a VT220 emulator. This would require that it provide the same escape sequences for the editing and numeric keypads. I could not simply change the escape sequences from "PC-style" to "VT220-style", since a number of users "knew" that the keypad "ought to" send home, end, cursor keys, etc., because they had labels indicating that use. To retain compatibility (but allow easy reconfiguration to make a VT220 emulator), I added popup-menu items to switch between the modes. With minor refinements, this was the approach for about two years, culminating with the "stable" patch #88, which is essentially the version distributed with XFree86 3.3.x.

NOTE:
the terminfo distributed with xterm patch #88 is incorrect: the escape sequences given for home/end keys are the VT220-style, rather than the default PC-style. Too accustomed to switching modes on the fly, I overlooked a line in my .Xdefaults file:

		*sunKeyboard: true
Downstream packagers (when they noticed this) accommodated the bug by modifying the VT100 translations resource which is not a good technical solution since it interferes with the users' ability to modify that resource. See this for more discussion.
But xterm continues to evolve past the stable patch #88. The keyboard support was still unsatisfactory for two reasons: The former could be addressed by expanding the escape sequences sent by the PC-style function keys, while the latter was a VT100/VT220 design issue. I decided to redesign function-key support to separate the two styles of function keys better, but leaving the choice still controlled by the sunKeyboard resource. Partway through that, I was asked to do similar cleanup and redesign of the backspace and delete key handling, e.g., the ptyInitialErase resource. Because it is a redesign, I chose to not make the keyboard differences between the old and new xterms completely compatible. If you were to run both on the same system, one or the other would have some problems with the editing keypad or the backspace/delete keys which would be addressed by the popup-menu selections.

For example, at this time (2001/9/4):

Here is a resource file which I tested with xterm-88c, xterm-149 and xterm-158, using $TERM set to xterm-debian.

Why can't I use the cursor keys in (whatever) shell?

VTxxx (VT100 and up) terminals may send different escape sequences for the cursor (arrow) keys depending on how they are set up. The choices are referred to as the normal and application modes. Initially, the terminal is in normal mode.

VTxxx terminals are usually set up so that full-screen applications will use the cursor application mode strings. This is good for full-screen applications, including legacy applications which may have hard-coded behavior, but bad for interactive shells (e.g., ksh, tcsh, bash) which use arrow keys to scroll through a history of command strings.

To see the difference between normal/application modes, consider this example:

Since termcaps and terminfo descriptions are written for full-screen applications, shells and similar programs often rely on built-in tables of escape sequences which they use instead. Defining keys in terms of the termcap/terminfo entry (e.g., by capturing the string sent by tputs) is apt to confuse the shell.

Depending on the terminal type, the keypad(s) on the keyboard may switch modes along with the cursor keys, or have their own independent modes. The control sequences for these are independent of the ones used for cursor-addressing, but are grouped together, e.g., as the terminfo smkx and rmkx capabilities. Terminfo entries are written assuming that the application has initialized the terminal using the smkx string before it is able to match the codes given for the cursor or keypad keys.

Why can't I cut/paste in xterm?

When an application sets xterm to any of its mouse tracking modes, it reserves the unshifted mouse button clicks for the application's use. Unless you have modified the treatment of the shifted mouse button events (e.g., with your window manager), you can always do cut/paste by pressing the shift key while clicking with the mouse.

Why can't I select tabs in xterm?

This issue was noted early on, here in 1997.

xterm is copying from the screen, which stores only printable characters. That includes spaces and line-drawing characters. But tabs are special; they are used for more than one purpose.

If the screen is cleared in some part, that stores nulls. Cursor addressing does not fill in nulls as it jumps around, though xterm does supply blanks for the most useful cases, especially when getting data for a selection.

Full-screen programs such as text-editors tend to write in random fashion, and generally do not print nulls to the screen. Curses on the other hand, may supply tabs where you thought there were none. Also, the terminal driver can expand tabs (and often is set to do this by default).

So the whole thing is unreliable: unless you make special arrangements for each of the programs running inside xterm, you would often get a tab when you expect, and vice versa.

For the special case where your expectations would match the available data, it is solvable. There are basically two ways it could be done:

As of 2010, a few other terminals do implement this feature. But the reason that it's been low-priority is that it's of very limited usefulness when copying between terminal sessions (and for that matter, from other clients).

Why does $LD_LIBRARY_PATH get reset?

If xterm is running setuid (which is needed on some systems which have no wrappers for opening pty's and updating utmp), newer systems automatically set or reset environment variables which are considered security problems. These include $PATH and $LD_LIBRARY_PATH, since they affect the choice of which programs are run if not specified via a full pathname.

This means, for example, that if you attempt to run


	xterm -e foo
where foo is a program that uses shared libraries in /usr/local/lib, then the command will fail, because /usr/local/lib is not considered part of root's environment.

Modern Unix systems (such as recent Solaris and HPUX versions) do not require you to run xterm setuid. Some will result in odd malfunctions if you do this.

Why do the -e and -ls options not work together?

Xterm has two useful options for controlling the shell that is run:
-e
tells xterm to execute a command using the remaining parameters after this option.
-ls
tells xterm to invoke a login shell, making it read your .login file, for instance.
The two are not compatible. If you specify both, xterm uses -e, and if that fails for whatever reason will fall through to the -ls option. It cannot (in general) combine the two, since some shells permit this (e.g., bash), and others do not (e.g., tcsh).

I need /etc/termcap

If you have a termcap version of xterm on a system with no termcap libraries, you may also be missing /etc/termcap.

A workaround is to copy /usr/X11R6/lib/X11/etc/xterm.termcap to /etc/termcap.

This is fixed another way starting with XFree86 3.3.1. If xterm cannot find the terminal description, it will accept that, though it will print a warning. If xterm does not find the termcap entry, it will not set the $TERMCAP variable.

Xterm does not run (no available pty's)

Your copy of xterm may not have enough permissions to use existing pty's:

Perhaps your system does not have enough pty's, or (problems reported with newer Linux kernels supporting Unix98 pty's, beginning with RedHat 6.0) the major device numbers of the pty's may have changed during a kernel upgrade. (This is described in /usr/src/linux/Documentation).

See also the MAKEDEV script, which usually exists under /dev.

Reverse video is not reset

When running less or other programs that do highlighting, you see the highlighting not turned off properly.

This may be due to incompatible terminal descriptions for xterm. With XFree86 3.2, I modified the terminal description for XFree86 xterm to use the VT220 (aka ISO 6429) controls that allow an application to turn off highlighting (or bold, underline) without modifying the other attributes. The X Consortium xterm does not recognize these controls.

If, for example, you are running an older xterm and rlogin to a system where the newer xterm has been installed, you will have this problem, because both programs default to $TERM set to xterm. The solution for mixed systems is to install the newer terminal description as as a different name (e.g., xterm-color) and set the termName resource accordingly in the app-defaults file for the system which has the newer xterm.

However – see below.

My colors changed in vim

Some vim users may notice their colors change after updating to patch 238. Before, some text would display in a dark color using a bold font. Now, it displays in a bright color and normal font.

This is not a bug, but the result of a feature tcap-query which was added for vim in 2000. Several vim users requested that it be enabled by default in the configure script. It allows vim to ask what characters the different function keys actually send, eliminating the chance that the termcap does not match.

Vim also asks how many colors the terminal supports. Since patch 148, xterm has responded with the number of distinct colors that it can display. By default, that is 16 (8 ANSI colors with bright counterparts for displaying PC-style "bold" text).

The interpretation of this depends on the application: termcaps do not tell how to display more than 8 colors. But vim understands how to tell xterm to display using 16 colors. It makes a difference when displaying bright colors. Vim has a table of 16 color names ("dos-colors"), which one can use to define parts of the color scheme. If the terminal supports only 8 colors (colors 0-7), vim uses the bold attribute to simulate colors 8-15.

Changing the color scheme to use bold where it is wanted will make the colors work as before – and work consistently with other terminals.

"grep --color" does not show the right output

GNU grep (version 2.5) introduced a --color option.

It does this for each highlighted match:

  1. it writes the text up to (not including the match)
  2. it writes an ANSI color control control sequence
  3. it writes the matched text
  4. it writes a control sequence to clear to the end of the line
  5. it writes an ANSI control sequence to reset graphic rendition.
  6. repeat this process until the entire line is written.

One problem is in the second and fourth steps. If the preceding text brought us up to the last column, then xterm (and any VT100-compatible terminal) is waiting for graphic text to wrap to the next line. Any controls would take effect on the current column position. Newlines are ignored while in this state.

However, if xterm gets a control sequence while waiting to wrap to the next line, it will update the screen according to that control. Then it is ready to accept more data. But at this point, it is no longer waiting to wrap; the special case is for newline versus graphic characters. For instance, backspacing clears the state (vttest illustrates this). So the data starts to write at the current column (the last one on the line), rather than at the beginning of the next line. In that case, grep's output will not look right.

What $TERM should I use?

The xterm-color value for $TERM is a bad choice for XFree86 xterm because it is commonly used for a terminfo entry which happens to not support bce. Complicating matters, FreeBSD (after dithering for a few years on the matter) introduced a bastardized version which implies the opposite sense of bce, (because it uses SGR 39 and 49), but does not set it. After lengthy discussion, FreeBSD began using the terminal descriptions which I've written.

The most recent XFree86 version's terminal description corresponds to xterm-xfree86 (also distributed with ncurses). I have continued to make changes; the most recent version is simply named xterm-new (also distributed with ncurses).

The term "bce" stands for "back color erase". Terminals such as XFree86 xterm and rxvt implement back color erase, others such as dtterm do not. (Roughly half of the emulators that I know about implement bce). When an application clears the screen, a terminal that implements back color erase will retain the last-set background color. A terminal that does not implement back color erase will reset the background color to the default or initial colors. Applications that paint most of the screen in a single color are more efficient on terminals that support back color erase.

Curses libraries that support color know about bce and do the right thing – provided that you tell them what the terminal does. That is the whole point of setting $TERM. The "xterm-color" description distributed with ncurses does not list bce, because it was applied originally to a terminal type which does not implement back color erase. It will "work" for XFree86 xterm, though less efficient. Some other applications such as the slang library have hardcoded support for terminals that implement back color erase. Given the "xterm-color" description, those will be efficient – and fortuitously work. However, slang (through version 1.4.0) did not work properly for the terminals that xterm-color was designed for. See this page for an example of (n)curses and slang running on dtterm. That bug in slang is reported to be fixed for succeeding versions, though your application may require changes to use this fix. (The demo which comes with slang to illustrate the use of bce does not work properly, for instance).

The xterm-color value for $TERM is also (for the same reason) a bad choice for rxvt, but "works" due to the large number of hard-coded applications that override this.

Some people recommend using xtermc. That is installed on Solaris. However, it does not match any xterm in current use. (Apparently it was written for an obsolete version on Unixware). The colors work, true, but the mouse will not, nor will the function keys.

FVWM does weird things when I try to resize xterm

I have an old (3.1.2G) bug report for xterm which may be related to the second (3.9s) problem: I have not observed the first, but have reproduced the second.

Why doesn't the screen clear when running vi?

This refers to the "alternate screen" feature, which has been used in its termcap file since 1988. On various systems, this feature may have been removed, although it has always been in the xterm sources.

The feature is controllable (it can be enabled or disabled). However, as it was originally conceived, that ability to control it applies only to programs using termcap.

Under SunOS 4.x, the termcap description for xterm embeds in the ti and te capabilities a command to switch to xterm's alternate screen (e.g., while running vi), and return to the normal screen on exit. This has the effect of clearing the screen. Under Solaris 2.x, the terminfo description does not use the alternate screen (it is a matter of preference after all), so that the text from vi remains on the screen after exit. There are corresponding terminfo symbols for ti and te: smcup and rmcup, respectively.

This is configurable...

For example (from Bjorn Helgaas <helgaas@dhc.net>) this procedure adds these capabilities to the "xterm" terminfo definition on HP-UX 10.20:


	cp /usr/lib/terminfo/x/xterm /usr/lib/terminfo/x/xterm.orig
	untic xterm > /tmp/xterm.src
	echo " smcup=\E7\E[?47h, rmcup=\E[2J\E[?47l\E8," >> /tmp/xterm.src
	tic /tmp/xterm.src
In this example, the terminfo strings are a series of operations:
smcup
\E7 saves the cursor's position
\E[?47h switches to the alternate screen
rmcup
\E[2J clears the screen (assumed to be the alternate screen)
\E[?47l switches back to the normal screen
\E8 restores the cursor's position.

However, xterms that are linked with termcap are more flexible in this area than those linked with terminfo libraries. The xterm program supports a resource titeInhibit which manipulates the $TERMCAP variable to accomplish this. It sets the $TERMCAP variable for the client with the ti and te capabilities suppressed. Systems that use terminfo cannot do this. If you are running terminfo with the alternate screen controls in the terminal description, then you can suppress the switching to the alternate screen by the titeInhibit, but not the associated cursor save/restore and clear-screen operations.

XFree86 3.9s xterm implements a different set of controls (private setmodes 1047, 1048 and 1049) which address this (in addition to the older set of controls, for compatibility). The new set of controls implements the entire ti sequence (save cursor, switch to alternate screen, clear screen) and te (switch to normal screen, restore cursor) as two control sequences that can be disabled by titeInhibit.

The 1049 code is a refinement of 1047 and 1048, clearing the alternate screen before switching to it rather than after switching back to the normal screen. Since patch #90 in 1998 xterm allows you (with a popup menu entry designed to exploit this behavior) to switch the display back to the alternate screen to select text from it, to paste into the normal screen. You can also set or clear the titeInhibit resource using another popup menu entry (Enable Alternate Screen Switching).

Most other terminal emulators implement only half of the feature. They recognize the control sequence, but do not provide the ability to change it at runtime, e.g., using a menu entry. Like any other half-done implementation, that is a bug which should be reported to the developers of those programs.

Why is the cursor misplaced after running vi?

Vi and other full-screen applications use the termcap ti/te (terminfo smcup/rmcup) strings to initiate and end cursor addressing mode. As mentioned in the discussion of titeInhibit, full-screen applications can expect the initialization string to save the cursor's position, and the end-string to restore it.

A few applications (reportedly IRIX 5.x and 6.x vi incorrectly move the cursor before initializing cursor-addressing. This will cause the end-string to restore the cursor to its position when it was saved by the initialization string (typically at the upper left corner of the screen).

The usual reason is due to the cursor save/restore controls in the ti/te strings. If your application runs a subprocess which in turn runs another full-screen application (or when reinitializing the screen after the shell process), it will save the cursor position again, so the position which is restored when finally exiting your program is the last one saved, not the first. XFree86 xterm (from late 1998, patch 90) changes the behavior of the cursor save/restore operations so they apply only to the current screen. That makes it less likely to misplace your cursor.

Why doesn't the scrollbar work?

Originally xterm was built using imake rather than a configure script. One feature of imake that is not possible to guess within the configure script is the wide-prototype compile-time definition NARROWPROTO. When this is not set properly, the Athena widget scrollbars do not work properly. xterm's configure script has a fallback case which allows disabling imake. However, this is moot with the Xorg "modular" build, whose compiler options are unrelated to imake or older versions of any libraries that it may distribute. In this case, the configure script needs some help. Use this option to enable or disable NARROW proto (and disable imake with the --disable-imake option) to match the whims of Xorg hackers.

For instance


	configure --disable-imake --disable-narrowproto

Why can't my program read the window title?

The longstanding control sequence for reading the window title is something that can be abused in special conditions. For novice (unknowledgable) users, this can be a problem.

xterm provides resource-settings and menu entries to allow this and related features to be enabled or disabled. See for example allowWindowOps The default resource settings in xterm can be overridden by a packager. However, a knowledgable user can override those default settings.

It is also possible that an overzealous packager may have crippled xterm by removing the functionality altogether. (That should be reported as a bug, to me).

For instance, one of those sent me a "security fix" some years ago, which deleted most of the control sequences which return data to the host. It broke the resize program, and selection, among other uses considered to be benign. In contrast, the same features used in other terminal emulators are tolerated by the same people, so rather than being a misguided attempt at fixing security issues, patches such as that appear to be an attempt at harassment.

Why can't my program set the window size?

Some overzealous packagers, perhaps influenced by the demonstration I provided, are protecting you against the possibility of your xterm becoming inaccessible. (That's unlikely...).

You should be able to override it, as noted above via resource settings or menu entry ("Allow Window Ops").

Why doesn't my delete key work?

This seems to be a problem with the older XFree86 release (3.1.2). I have picked up pieces of the story (xterm and the keyboard work as designed under XFree86 3.2 and up).

The underlying problem is that we've accumulated three things that are being equated as "Delete":

	ASCII backspace (code 8)
	ASCII delete (code 127)
	VT220 "remove" aka "delete" (ESC [ 3 ~)

You are probably talking about the backarrow key (on my keyboard, at the upper right of the QWERTY block), or the key labeled delete which is on the 6-key "editing keypad". Since xterm is emulating a VT100/VT220, the backarrow key should generate a 127 (often displayed as ^?). You would use a control/H to obtain a backspace on a real VT220.

Tastes differ on Unix, people expect the backarrow key to generate a backspace (or not). As I understand it, at one point, XFree86 picked up the sense of the erase character during initialization, so that xterm would in effect use the same erase character as the console. The current scheme (X11R6) uses keyboard mapping tables that are independent of the environment.

XFree86 xterm provides a resource toggle backarrowKey (and an escape sequence from VT320) that changes this key between the two styles (backspace or delete).

With XFree86 xterm patch 95 (also in the stable version as "88c"), you may have an xterm which can automatically initialize the backarrow key to backspace or delete depending on the pseudo terminal's sense, or based on the termcap setting of kbs (backspace key). This feature is controlled by the resource setting ptyInitialErase.

Why did my delete key stop working?

Well, something changed. You have to determine what did.

This may be because an upgrade introduced different X resource settings, or because you are using the newer xterm with the ptyInitialErase resource (or perhaps both). Use


	appres XTerm
to see the X resources that you are using, in particular the translation (or Translation) resource for the vt100 widget.

One unexpected scenario came out of hiding when I was implementing the ptyInitialErase resource. When xterm is (by default) built to support this, it sets the pty's erase character to match the termcap entry. Xterm also sets the $TERMCAP environment variable to match. So everything is consistent, and everything defined. The stty erase character is either backspace (^H) or delete (^?).

The problem arises because there are two things called "delete", which were not well-defined: ASCII delete (127) and the PC-style adaptation of VT220 remove assigned to the key Delete.

However, the screen program prefers to make the termcap delete (kD) an <escape>[3~, which corresponds to the VT220 remove key. If $TERMCAP is set when starting screen, it will translate stty's erase character into the <escape>[3~, making most curses and termcap applications work. But stty still has the original erase character. So low-level applications which check stty will not work. I found that unsetting $TERMCAP before running would work, but this was not a good solution. Someone pointed out (see patch 129), that the problem really was because termcap kD should delete the character at the current position. So it cannot be the same as stty erase.

As a matter of fact, stty erase has to be a single character, so <escape>[3~ would not work anyway.

Well, how can I set my delete key?

When people first started asking this question in 1995-1996, it appeared in the context of making Netscape work. Netscape's use of the delete key running in X did not match user's expectations when compared to that other platform. They were commonly advised to use xmodmap, e.g.,

	keysym BackSpace = Delete
or

	keycode 22 = 0xff08
Either way is a bad technical solution – it works for some people but not others (on my keyboard at work, keycode 22 is the numeric keypad '9').

Alternatively, you can set resources. This works reasonably well for environments where you have different versions of xterm, e.g.,


	XTerm*VT100.translations: #override \n\
		<Key>Delete: string(0x7f)
I do not do that either, because it is not flexible. Not all programs use the same sense of stty erase; some use termcap or terminfo, and some are hardcoded. So I prefer to be able to switch the xterm's keyboard at runtime. You cannot do that with resources. (Or not really – xterm has a keymap() action which could support this if you provided a rather complex resource settings, but the X library support for that is broken in X11R6). Instead, I have added to XFree86 a set of resources (and popup menu entries) to allow simple switching between the different styles of keyboard, in particular for the backspace/delete issues. See the manual page for backarrowKey backarrowKeyIsErase and deleteIsDEL as well as sunKeyboard.

Why doesn't my keypad work?

A few people have commented that the keypad does not work properly. Aside from bugs (I have fixed a few), the most common problem seems to be misconception.

Here's a picture of the VT100 numeric keypad:

	+-----+-----+-----+-----+
	| PF1 | PF2 | PF3 | PF4 |
	+-----+-----+-----+-----+
	|  7  |  8  |  9  |  -  |
	+-----+-----+-----+-----+
	|  4  |  5  |  6  |  ,  |
	+-----+-----+-----+-----+
	|  1  |  2  |  3  |     |
	+-----+-----+-----+ ENT +
	|     0     |  .  |     |
	+-----+-----+-----+-----+
and the similar Sun and PC keypads:
	+-----+-----+-----+-----+
	| NUM |  /  |  *  |  -  |
	+-----+-----+-----+-----+
	|  7  |  8  |  9  |     |
	+-----+-----+-----+  +  +
	|  4  |  5  |  6  |     |
	+-----+-----+-----+-----+
	|  1  |  2  |  3  |     |
	+-----+-----+-----+ ENT +
	|     0     |  .  |     |
	+-----+-----+-----+-----+
Working in X11, the NUM (NumLock) key has better uses than an alias for PF1 (and is sometimes reserved). I use the F1 through F4 on the keyboard to implement PF1 through PF4, alias the keypad "+" to "," and use the existing "-" key.

VT220 emulation uses the VT100 numeric keypad as well as a 6-key editing keypad. Here's a picture of the VT220 numeric keypad:

	+--------+--------+--------+
	| Find   | Insert | Remove |
	+--------+--------+--------+
	| Select | Prev   | Next   |
	+--------+--------+--------+
and the similar Sun and PC keypads:
	+--------+--------+--------+
	| Insert | Home   | PageUp |
	+--------+--------+--------+
	| Delete | End    | PageDn |
	+--------+--------+--------+

I chose to use keys that are mnemonic rather than in the "same" positions, though some emulators (e.g., Tera Term) use the same positions:

	+--------+--------+--------+
	| Insert | Find   | Prev   |
	+--------+--------+--------+
	| Remove | Select | Next   |
	+--------+--------+--------+

I test the keyboard (for VT52/VT100/VT220) using vttest. If you find (or think that you have found) a problem with the keyboard handling of xterm, please test it with vttest first.

Other arrangements of the keyboard are possible of course. If you prefer to use the top row of the numeric keypad as PF1 through PF4, you should do this using xterm's X resources.

Why can't I input 8-bit characters?

You must have the eightBitInput resource set to do this.

My .Xdefaults for XTERM


	XTerm*internalBorder:  10
	XTerm*highlightSelection:  true
	XTerm*VT100.colorBDMode:  on
	XTerm*VT100.colorBD:  blue
	XTerm*VT100.colorULMode:  on
	XTerm*VT100.colorUL:  magenta
	XTerm*VT100.eightBitInput:  true
	XTerm*VT100.eightBitOutput:  true
	XTerm*scrollBar:  true
	XTerm*VT100.titeInhibit:  true

	XTerm*VT100.colorMode:  on
	XTerm*VT100.dynamicColors:  on

	! Uncomment this to use color for underline attribute
	XTerm*VT100.colorULMode:  on
	XTerm*VT100.underLine:  off

	! Uncomment this to use color for the bold attribute
	XTerm*VT100.colorBDMode: on

	XTerm*VT100.color0: black
	XTerm*VT100.color1: red3
	XTerm*VT100.color2: green3
	XTerm*VT100.color3: yellow3
	XTerm*VT100.color4: blue3
	XTerm*VT100.color5: magenta3
	XTerm*VT100.color6: cyan3
	XTerm*VT100.color7: gray90
	XTerm*VT100.color8: gray30
	XTerm*VT100.color9: red
	XTerm*VT100.color10: green
	XTerm*VT100.color11: yellow
	XTerm*VT100.color12: blue
	XTerm*VT100.color13: magenta
	XTerm*VT100.color14: cyan
	XTerm*VT100.color15: white
	XTerm*VT100.colorUL: yellow
	XTerm*VT100.colorBD: white

	XTerm*VT100.cursorColor: lime green

XFree86 xterm comes with two copies of the resource file, one with color only (XTerm-col.ad, which is installed as XTerm-color), and the regular one (XTerm.ad, installed as XTerm). To use the XTerm-color file in conjunction with a separate XTerm app-defaults file which does not contain color, add the following line to your .Xdefaults file:

	*customization: -color

Why are the menus tiny?

Everything seems to work, except that the xterm menus (VT options, fonts, etc.) do not display properly; the menus pop up, but only with a tiny display area in which none of the options are visible (and only part of the menu title is visible).

You have specified the geometry for xterm too high in the hierarchy, and that 24x80 (or whatever the -geometry parameter happens to be) is applying to the menus in pixels. This resource makes the geometry apply to the menus as well as the VT100 widget:


	XTerm*geometry: 80x24
while this applies only to the VT100 widget (which is probably what you intended):

	XTerm.VT100.geometry: 80x24
or better yet (to allow for the toolbar option, which uses a level of widget hierarchy):

	XTerm*VT100.geometry: 80x24

What is this warning message?

xterm: Error 11, errno 22: permission denied
Actually, any message like this denotes a failure which requires studying the xterm source to determine the exact problem.

You have either found a bug in xterm, or there is something wrong with your computer's configuration, e.g., not enough pty's, incorrect permissions, etc.

The first number is an internal code (defined in error.h in xterm's source), and the second is the system error number (defined in /usr/include/sys/errno.h). The system error number is easier to lookup, but the internal error code tells you where to look in the source.

input method doesn't support my preedit type
Ignore this if you do not know what input method is. Input methods are used to enter composite characters (e.g., umlauts, other types of punctuated characters, East Asian characters, etc). Your computer's libraries support this, but are missing configuration tables, and xterm is warning you.

If the message bothers you (e.g., if you aren't starting xterm from a window manager menu), you can suppress it by setting a resource:


		XTerm*openIm:false
Warning: Actions not found: ignore, "xxx"
The action "xxx" (for example "scroll-back") is specified in a resource file whose translations match widgets that do not support them. For example, this

		XTerm*translations:     #override\n\
			<Leave>, ~Ctrl ~Meta <Btn2Up>: ignore()\n\
			~Shift <Key>KP_8: scroll-back(1,line)\n\
			~Shift <Key>KP_2: scroll-forw(1,line)\n\
			Shift <Key>KP_8: scroll-back(1,halfpage)\n\
			Shift <Key>KP_2: scroll-forw(1,halfpage)
will produce warnings such as

		Warning: Actions not found: ignore, scroll-back, scroll-forw
		Warning: Actions not found: ignore, scroll-back, scroll-forw
		Warning: Actions not found: ignore, scroll-back, scroll-forw
This is a correct form, assigning the actions to the "VT100" widget.

		XTerm*VT100.translations:     #override\n\
			<Leave>, ~Ctrl ~Meta <Btn2Up>: ignore()\n\
			~Shift <Key>KP_8: scroll-back(1,line)\n\
			~Shift <Key>KP_2: scroll-forw(1,line)\n\
			Shift <Key>KP_8: scroll-back(1,halfpage)\n\
			Shift <Key>KP_2: scroll-forw(1,halfpage)
Warning: Cannot allocate colormap entry for "xxx"
This comes from the X library. XFree86 xterm uses the default color map. What this means is that if your X server has insufficient space to store color information for more than one color map, other applications which could use other color maps may conflict with xterm. In practice, that is 256 unique colors on the screen at a time -- not enough for a fancy background or an application such as Netscape.

During resource initialization, xterm attempts to allocate an entry from the color map for each color which it might use. If there are not enough free slots in the color map, you will see a "Cannot allocate" message for each color that xterm failed to allocate. Those colors will be rendered in the foreground color, making full-screen color applications such as dialog unreadable.

This problem is alleviated with patch 129, which modified xterm to delay the most color allocation until the colors are first needed. If a color is never needed (xterm allocates 20 colors in this manner), that reduces the number of slots in the color map that are needed. Even with this improvement, xterm must still allocate 4 colors during initialization to determine how to display the cursor. If none of those colors can be allocated, XFree86 xterm reverts to monochrome.

Known Bugs in XTERM

These are the known bugs (or limitations) in the XFree86 xterm. They are also present in the other versions based on the X Consortium sources (color_xterm, ansi_xterm, kterm).

Note that of the emulators that support color, most do not support bce (back color erase). The bce capability is also called the "new color model", though it has been implemented in the IBM PC for quite a while. Technically, not implementing bce (or allowing the choice between it and its complement) is not a bug, since few hardware terminals (with good reason) implemented this feature.

X11R6.3 XTERM

The X Consortium version of xterm (and versions based on it) has additional bugs not in XFree86 xterm: (These bugs are also present in the X11R5 version).

Update 2004/04/08:
Complicating this discussion is the "X.Org" xterm (from 2004). That is the XFree86 xterm from XFree86 CVS with all visible "xfree86" strings changed to "X.Org" or "xorg", depending on the use. For example the "xterm-xfree86" terminfo entry becomes "xterm-xorg". The change history for the related CVS for X.Org shows this. Similarly, the release notes for X11R6.7 included my notes for XFree86 4.4.

As of 2009, it appears that "X.Org" xterm died a natural death, since none of the people who created it had any likelihood of maintaining it. Instead, X.Org defers to my version of xterm.

COLOR_XTERM download

This is based on the X Consortium X11R5 source, with the same bugs. Not exactly a bug, but it does not build on Linux with X11R6.3

ANSI_XTERM download

This is based on the X Consortium source, with the same bugs.

CXTERM download

CXterm stands for "Chinese Xterm". This is based on the X Consortium source.

DTTERM

This is distributed with CDE. It implements more of the DEC VT220 than the X Consortium xterm, and also adds controls to manipulate the window and icon.

EMU 1.3 download

This is not based on the X Consortium source. The authors state that it implements VT220 emulation. It is in need of maintenance, since it builds with some problems to produce an executable that (on Linux and SunOS) does not handle the carriage return and newline translations properly. So I am unable to run vttest on this emulator.

ETERM link

Eterm was based on rxvt, though the appearance differs. The terminal emulation capabilities appear similar, though I am not able to run the full suite of tests in vttest with this emulator (the core dump noted for rxvt, as well as hanging while awaiting response from one or more control sequences). Oddly, it appears that neither Eterm nor rxvt implement CPR (cursor position report). Finally, it reserves F1 (function-key) for a popup menu. This applies to versions of Eterm through 0.9.

GNOME TERMINAL link

GNOME Terminal is developed separately from both xterm and rxvt, and is based on the zvt (zterm) widget. Like kvt), it appears to have been developed imitating other terminal emulators (Linux console and xterm) rather than strictly emulating a VT102. The documentation is fragmentary (with a comment suggesting that the author does not know where to find relevant information), and the program fares badly with vttest. Beginning with late 1999, reports indicate that it does not properly parse ANSI control sequences: the vim editor is using XFree86 xterm's vt220-style "Send Device Attributes" (Secondary DA) control sequence to obtain the terminal emulator's version. That is, it sends

	\E[>c
expecting a response such as

	\E[>0;138;0c
for vt100. The bug report indicates that the "c" sent by vim is echoed rather than interpreted by the emulator.

But it suffices for vi.

A more recent GNOME Terminal (version 1.4.0.4, late 2001) did not implement a complete vt102: it was missing several features which can be demonstrated in vttest). Most of the bugs in the Device Attributes responses remain, but it works a little better with vim. However, there are problems with the alternate screen that show up with vim. Again, these can be demonstrated with vttest (menu 11.6.3 in the 20011130 snapshot).

The documentation for GNOME terminal asserts:

GNOME Terminal emulates the xterm application developed by the X Consortium. In turn, the xterm application emulates the DEC VT102 terminal and also supports the DEC VT220 escape sequences. An escape sequence is a series of characters that starts with the Esc character. GNOME Terminal accepts all of the escape sequences that the VT102 and VT220 terminals use for functions such as to position the cursor and to clear the screen.

That sounds fine, except that it is both inaccurate and misleading:

inaccurate
combining the "X Consortium" and "DEC VT220", for example, since that was done after the demise of said organization.

It emulates a subset of VT100, lacks support for most of the VT220 control sequences (including some used for positioning the cursor) that are not recognized by a VT100.

Even in the subset which it emulates, GNOME Terminal has bugs. Many of these are easy to demonstrate with vttest.

misleading
as noted in Xterm Control Sequences, xterm (mostly after "X Consortium") supports control sequences which are not VT100/VT220. GNOME Terminal implements many of these, but not all.

Perhaps that was unintentional – GNOME developers did not appear to document what their program does outside of that remark. However, an inspection of the changelog for libvte (VTE) does show that most of the borrowing from xterm is cited in an oblique manner – not once mentioning XFree86 for example, leaving the impression (as indicated by "X Consortium") that all of the work on xterm was done before development of GNOME Terminal commenced.

Most of this observation was documented between 2000 and 2007. Other than maintenance, development of GNOME Terminal appears to have paused in 2005. As of 2009, its maintainer is (of the development team), the least knowledgeable about terminal emulation. So there is no progress on the large number of bug reports related to xterm-compatibility.

Notes on VTE

VTE's README file asserts

VTE supports Unicode and character set conversion, as well as emulating any terminal known to the system's terminfo database.

The latter part of that ("emulating any terminal") is incorrect. It does have the ability to work with the standard function-key definitions which can be defined in a terminfo description.

Some of the function-key logic is adapted from xterm; generally refactoring the xterm source-code to make it appear different. In places however (naming conventions and comments), there is some verbatim copying. The same comment is true of "character set conversion". None of that is reflected in VTE's changelog.

xterm on the other hand, can be told with the tcapFunctionKeys resource setting to use a more complete subset, based on the ncurses extended terminal descriptions. Even xterm's terminfo/termcap descriptions do not cover the (literally) thousands of keyboard combinations which are available via its resource settings.

Outside of function-keys, VTE provides no ability to emulate "any terminal". A casual glance at its source code reveals the following:

For instance, VTE cannot emulate dtterm, because of differences in color behavior. In fact, VTE does not use any of the termcap data to support its interpretation of color control sequences.

Because of GNOME Terminal's reputation for excessive code bloat, developers of every other program based on VTE advertise their version as reduced memory usage, faster startup, etc. Here are a few of the available ones:

osso-xterm link
This is one of several GUI "skins" over the VTE widget which forms the functional core of GNOME Terminal.

Its home page refers to "at least two versions". I recall seeing an older version which was apparently not based on VTE. There does not appear to be any current page (as of 2009) for that version.

roxterm link
This is one of several GUI "skins" over the VTE widget which forms the functional core of GNOME Terminal.
XFCE Terminal link
This is one of several GUI "skins" over the VTE widget which forms the functional core of GNOME Terminal.

MULTI GNOME TERMINAL (MGT) link

Of particular note, MGT 1.4.0 announcement claims that it works properly for all of vttest)'s tests. On the positive side, it does do VT52 emulation, but (reading the source code did not help) it apparently does not really do VT220 from vttest's perspective.

HANTERM download

HanTerm stands for "Hangul term" (Korean). This is based on the XFree86 source.

KONSOLE link

More than just a rewrite of kvt into C++. But there are several incompatibilities between konsole (noted with version 1.0.2 in late 2001) and xterm: The problems with setmode 1049 were fixed after some time; other issues linger on.

Like GNOME Terminal, konsole's documentation is incomplete and inaccurate. This gem from its handbook illustrates the problem:

After a decade, Konsole is the first rewrite from the ground up. While xterm has definitely been hacked to death (its README begins with the words Abandon All Hope, Ye Who Enter Here), Konsole offers a fresh start using contemporary technologies and understanding of X.
The problem:

KTERM download

KTerm stands for "Kanji term" (Japanese). This is based on the X Consortium source, with the same bugs (though the list of original authors has been removed; the modifications that comprise kterm is relatively small). There is a variation of xvt (ancestor of rxvt) originally known as kvt bundled with KDE which may be referred to as "kterm", but I do not find it interesting, other than to comment that it was a poor choice of name.

MLTERM

Mlterm is not based on xterm or rxvt source, though it implements many of their features. It does fairly well with vttest, except for some odd misbehavior in operations that save/restore the cursor position.

MTERM

There are a few variants of this: the xterm bundled with some Motif clients is more common. More interesting, however is one (not Motif), attributed to "Der Mouse".
(mouse@Lightning.McRCIM.McGill.EDU) Available: larry.mcrcim.mcgill.edu (132.206.1.1) in /X/mterm.src/mterm.ball-o-wax.
I saw only an incomplete version of this while it was advertised in the mid-90's. It is available by email from <mouse@Rodents.Montreal.QC.CA>. or via ftp. This is not a patched version of xterm, though it was apparently written, like rxvt, to emulate vt100's. While it does have some interesting features (such as blinking characters), overall it does not do as well with vttest) as the more widely known emulators.

MXTERM

There are several variants on this: xterm adapted for Motif libraries. I have seen none that work properly:

NXTERM

Distributed with Redhat Linux 5.2, it is a repackaging of xterm-sb_right-ansi, to use the Xaw3d widget set. This is based on the X Consortium X11R6 source, with the same bugs. Starting with Redhat 6.0, nxterm is the XFree86 3.3.6 xterm. Unfortunately Redhat neglected to update their termcap for nxterm to match the program.

RXVT link

Rxvt's manual page states the following unqualified comment:
rxvt, version 2.6.2, is a colour vt102 terminal emulator intended as an xterm(1) replacement for users who do not require features such as Tektronix 4014 emulation and toolkit-style configurability. As a result, rxvt uses much less swap space -- a significant advantage on a machine serving many X sessions.
How much is much less? Perhaps not as much as one would think from reading that. The Tektronix emulation in xterm (which has been optional since late 1997) accounts for about 25kb of the code.

The toolkit-style configurability glibly referenced is the ability to redefine keys on the keyboard without recompiling the program, i.e., the translations resource. It also is the way mouse events and other actions are passed to xterm.

The toolkit-style configurability accounts for about 300kb, which does add up if you happen to be running 50 xterm processes (i.e., about 10Mb).

Compared with something like GNOME Terminal, which takes 2-3 times, or KDE konsole, which takes 15-20 times as much memory to run, xterm and rxvt memory requirements are indistinguishable to the normal user.

These comments apply to versions of rxvt through 2.21:

A newer version (upgraded to an beta as of 2.6.PRE3, however, since it no longer dumps core in vttest) is reported to fix the ech bug. However, it is less VT100-compatible than the earlier versions such as 2.21b because it does not render reverse video (DECSCNM) properly. All versions do not update the screen frequently enough, making animation ineffective. See vttest, tests 1 and 2.

One longstanding issue with rxvt impacts use of xterm. While rxvt does not use the X Toolkit (and corresponding X resource matching), it does read your .Xdefaults and app-defaults files to extract resource settings. That in itself would not be a problem. However, since rxvt also looks for resources in the XTerm class (a parasitic relationship like setting $TERM to "xterm" based on the presumption that it is a nuisance to install its configuration files), there have been several occasions on which xterm's app-defaults files have been modified to accommodate rxvt's variant usage.

That comment applies mainly to the resource patterns. However, even when the pattern is reasonably unambiguous, but overbroad, the results can be conflicting. For example, some versions of rxvt may accept a font resource which does not match the XLFD pattern. It accepts a prefix of "xft:". This feature (apparently introduced by konsole) tells rxvt to interpret the remainder of the string as a TrueType (Xft) font rather than a bitmap font. xterm uses the faceName resource for these values.

XGTERM link

It has some features which are also in color_xterm:(non-bce ANSI color, colorBD and colorUL resources, cursor warping, etc. The main feature is its Tektronix graphics emulation, which is the main reason for this particular program. Neither program has a change-log, so it is not easy to say which influenced the other.

That is from reading the source code. However testing under Debian Linux, something is wrong with the resource processing (neither popup menus nor colors work).

XITERM link

This appears to be rxvt 2.20, lightly reformatted, with a few ifdef's changed.

That is, it was. The name was later appropriated by a different program, which also uses the name iterm. Like gnome-terminal, iterm aims to be an xterm-emulator rather than a VT102- or VT220-emulator.

An earlier attempt by the same author (the "CSI-xterm") incorporates some of the changes I made for XFree86 xterm via cut and paste (but does not mention this in its README). It is said to be the basis for Solaris 10 xterm.

Both have similar problems running vttest.

How do I build XTERM?

Building a copy of xterm is simple, provided that you have a development configuration for X11: If you have a working xmkmf script (or correctly configured imake utility), all you need to do is type

	xmkmf
	make
I have written a simple configure script for xterm which uses imake (or xmkmf) to generate a Makefile from the Makefile.in. I have restructured xterm to eliminate most hardcoded #ifdef's, replacing them with definitions that can be derived with the configuration script. The configure script is more flexible than xmkmf, since it allows you to enable or disable a variety of features. Type
	configure --help
to get a list of options.

Though I have replaced most hardcoded ifdef's with autoconfigured values, it will still continue to build properly with the imake environment.

How do I report bugs?

You should report bugs to me. I also respond to bug reports in a number of bug-tracking systems, though some are less open to searches than others. See also Analyzing problems with configure scripts

Additional Information

There appears to be no comprehensive source of information on xterm better than the documentation which comes with the source code.

Other Sites

I have found Richard Shuford's archive to be invaluable for notes on the DEC VT220 and related terminals. Though not available at the time that I was collecting most of my notes, VT100.net is also a good source of primary information.

The XTerm Manual

The command-line options, X resources and similar configurable options of xterm are documented in the manual page.

Here are copies of the file in various forms: html, pdf and text.

Xterm Control Sequences

Control sequences, i.e., programming information are in the ctlseqs.ms file which I bundle with the program source. (It used to be in the same directory in the X distribution, but was moved to a different part of the tree long ago). Note that you must format this file with different options than a manpage, e.g.,

	tbl ctlseqs.ms | nroff -ms >ctlseqs.txt
	tbl ctlseqs.ms | groff -ms >ctlseqs.ps
As a PostScript or PDF file, the individual letters of the control sequences are all boxed, for emphasis, but I find the text file equally readable.

Here are copies of the file in various forms: html, pdf and text.

resize – set TERMCAP and terminal settings to current xterm window size

resize is useful by itself, but is maintained for historical reasons as part of xterm. html, pdf and text.

luit – Locale and ISO 2022 support for Unicode terminals

luit also is maintained as part of xterm, since its upstream maintainer is inactive, and the ostensible maintainers have more than once delivered unusable versions, causing many bug reports to be issued against xterm.

Ongoing/future work