First, tset retrieves the current terminal mode settings
for your terminal.
It does this by successively testing
.IP • 4
the standard error,
.IP • 4
standard output,
.IP • 4
standard input and
.IP • 4
ultimately ``/dev/tty''
to obtain terminal settings. Having retrieved these settings, tset remembers which file descriptor to use when updating settings.
Next, tset determines the type of terminal that you are using. This determination is done as follows, using the first terminal type found.
1. The terminal argument specified on the command line.
2. The value of the TERM environmental variable.
3. (BSD systems only.) The terminal type associated with the standard error output device in the /etc/ttys file. (On System-V-like UNIXes and systems using that convention, getty does this job by setting TERM according to the type passed to it by /etc/inittab.)
4. The default terminal type, ``unknown''.
If the terminal type was not specified on the command-line, the -m option mappings are then applied (see the section TERMINAL TYPE MAPPING for more information). Then, if the terminal type begins with a question mark (``?''), the user is prompted for confirmation of the terminal type. An empty response confirms the type, or, another type can be entered to specify a new type. Once the terminal type has been determined, the terminal description for the terminal is retrieved. If no terminal description is found for the type, the user is prompted for another terminal type.
Once the terminal description is retrieved,
.IP • 4
if the ``-w'' option is enabled, tset may update
the terminal's window size.
When invoked as reset, tset sets the terminal
modes to ``sane'' values:
.IP • 4
sets cooked and echo modes,
.IP • 4
turns off cbreak and raw modes,
.IP • 4
turns on newline translation and
.IP • 4
resets any unset special characters to their default values
before doing the terminal initialization described above. Also, rather than using the terminal initialization strings, it uses the terminal reset strings.
The reset command is useful
after a program dies leaving a terminal in an abnormal state:
.IP • 4
you may have to type
<LF>reset<LF>
(the line-feed character is normally control-J) to get the terminal
to work, as carriage-return may no longer work in the abnormal state.
.IP • 4
Also, the terminal will often not echo the command.
The options are as follows:
The arguments for the -e, -i, and -k options may either be entered as actual characters or by using the ``hat'' notation, i.e., control-h may be specified as ``^H'' or ``^h''.
If neither -c or -w is given, both options are assumed.
When the -s option is specified, the commands to enter the information into the shell's environment are written to the standard output. If the SHELL environmental variable ends in ``csh'', the commands are for csh, otherwise, they are for sh. Note, the csh commands set and unset the shell variable noglob, leaving it unset. The following line in the .login or .profile files will initialize the environment correctly:
The -m options maps from some set of conditions to a terminal type, that is, to tell tset ``If I'm on this port at a particular speed, guess that I'm on that kind of terminal''.
The argument to the -m option consists of an optional port type, an optional operator, an optional baud rate specification, an optional colon (``:'') character and a terminal type. The port type is a string (delimited by either the operator or the colon character). The operator may be any combination of ``>'', ``<'', ``@'', and ``!''; ``>'' means greater than, ``<'' means less than, ``@'' means equal to and ``!'' inverts the sense of the test. The baud rate is specified as a number and is compared with the speed of the standard error output (which should be the control terminal). The terminal type is a string.
If the terminal type is not specified on the command line, the -m mappings are applied to the terminal type. If the port type and baud rate match the mapping, the terminal type specified in the mapping replaces the current type. If more than one mapping is specified, the first applicable mapping is used.
For example, consider the following mapping: dialup>9600:vt100. The port type is dialup , the operator is >, the baud rate specification is 9600, and the terminal type is vt100. The result of this mapping is to specify that if the terminal type is dialup, and the baud rate is greater than 9600 baud, a terminal type of vt100 will be used.
If no baud rate is specified, the terminal type will match any baud rate. If no port type is specified, the terminal type will match any port type. For example, -m dialup:vt100 -m :?xterm will cause any dialup port, regardless of baud rate, to match the terminal type vt100, and any non-dialup port type to match the terminal type ?xterm. Note, because of the leading question mark, the user will be queried on a default port as to whether they are actually using an xterm terminal.
No whitespace characters are permitted in the -m option argument. Also, to avoid problems with meta-characters, it is suggested that the entire -m option argument be placed within single quote characters, and that csh users insert a backslash character (``\'') before any exclamation marks (``!'').
A reset command appeared in 2BSD (April 1979), written by Kurt Shoens. This program set the erase and kill characters to ^H (backspace) and @ respectively. Mark Horton improved that in 3BSD (October 1979), adding intr, quit, start/stop and eof characters as well as changing the program to avoid modifying any user settings.
Later in 4.1BSD (December 1980), Mark Horton added a call to the tset program using the -I and -Q options, i.e., using that to improve the terminal modes. With those options, that version of reset did not use the termcap database.
A separate tset command was provided in 2BSD by Eric Allman. While the oldest published source (from 1979) provides both tset and reset, Allman's comments in the 2BSD source code indicate that he began work in October 1977, continuing development over the next few years.
In September 1980, Eric Allman modified tset, adding the code from the existing ``reset'' feature when tset was invoked as reset. Rather than simply copying the existing program, in this merged version, tset used the termcap database to do additional (re)initialization of the terminal. This version appeared in 4.1cBSD, late in 1982.
Other developers (e.g., Keith Bostic and Jim Bloom) continued to modify tset until 4.4BSD was released in 1993.
The ncurses implementation was lightly adapted from the 4.4BSD sources for a terminfo environment by Eric S. Raymond <esr@snark.thyrsus.com>.
Neither IEEE Std 1003.1/The Open Group Base Specifications Issue 7 (POSIX.1-2008) nor X/Open Curses Issue 7 documents tset or reset.
The AT&T tput utility (AIX, HPUX, Solaris) incorporated the terminal-mode manipulation as well as termcap-based features such as resetting tabstops from tset in BSD (4.1c), presumably with the intention of making tset obsolete. However, each of those systems still provides tset. In fact, the commonly-used reset utility is always an alias for tset.
The tset utility provides for backward-compatibility with BSD environments (under most modern UNIXes, /etc/inittab and getty(1) can set TERM appropriately for each dial-up line; this obviates what was tset's most important use). This implementation behaves like 4.4BSD tset, with a few exceptions specified here.
A few options are different
because the TERMCAP variable
is no longer supported under terminfo-based ncurses:
.IP • 4
The -S option of BSD tset no longer works;
it prints an error message to the standard error and dies.
.IP • 4
The -s option only sets TERM, not TERMCAP.
There was an undocumented 4.4BSD feature that invoking tset via a link named ``TSET'' (or via any other name beginning with an upper-case letter) set the terminal to use upper-case only. This feature has been omitted.
The -A, -E, -h, -u and -v options were deleted from the tset utility in 4.4BSD. None of them were documented in 4.3BSD and all are of limited utility at best. The -a, -d, and -p options are similarly not documented or useful, but were retained as they appear to be in widespread use. It is strongly recommended that any usage of these three options be changed to use the -m option instead. The -a, -d, and -p options are therefore omitted from the usage summary above.
Very old systems, e.g., 3BSD, used a different terminal driver which was replaced in 4BSD in the early 1980s. To accommodate these older systems, the 4BSD tset provided a -n option to specify that the new terminal driver should be used. This implementation does not provide that choice.
It is still permissible to specify the -e, -i, and -k options without arguments, although it is strongly recommended that such usage be fixed to explicitly specify the character.
As of 4.4BSD, executing tset as reset no longer implies the -Q option. Also, the interaction between the - option and the terminal argument in some historic implementations of tset has been removed.
The -c and -w options are not found in earlier implementations.
However, a different window size-change feature was provided in 4.4BSD.
.IP • 4
In 4.4BSD, tset uses the window size from the termcap description
to set the window size if tset is not able to obtain the window
size from the operating system.
.IP • 4
In ncurses, tset obtains the window size using
setupterm, which may be from
the operating system,
the LINES and COLUMNS environment variables or
the terminal description.
Obtaining the window size from the terminal description is common to both implementations, but considered obsolescent. Its only practical use is for hardware terminals. Generally speaking, a window size would be unset only if there were some problem obtaining the value from the operating system (and setupterm would still fail). For that reason, the LINES and COLUMNS environment variables may be useful for working around window-size problems. Those have the drawback that if the window is resized, those variables must be recomputed and reassigned. To do this more easily, use the resize(1) program.
This describes ncurses version 6.2 (patch 20200222).