Hi Alex & Helge, tl;dr: Learn and use the "-rU0" option when your terminal device doesn't support OSC 8 hyperlinks.
groff_man(7):
-rU0 Disable generation of URI hyperlinks in output drivers
capable of them, making the arguments to MT and UR calls
visible as formatted text. grohtml(1), gropdf(1), and
grotty(1) enable hyperlinks by default (the last only if
not in its legacy output mode).
If you _never_ use a terminal device that supports hyperlinking...
groff_man(7):
/.../share/groff/site-tmac/man.local
Put site‐local changes and customizations into this file.
Thus, Helge might add
.if n .nr U 0 \" Format URLs on terminal devices.
to his system's "man.local" file.
At 2025-08-25T21:55:28+0200, Alejandro Colomar wrote:
> > > > > That URI has '\\:' in it, which is correct in roff(7) (and in
> > > > > man(7)) source code. That is removed by troff(1) when
> > > > > formatting the page. If you read the formatted page that's
> > > > > not there.
> > > >
> > > > Yes, then no URL is there :))
> > >
> > > Hmmm, that depends on your terminal. If there's no URL or
> > > hyperlink, this might be an issue in either the terminal or
> > > groff(1).
> >
> > I use the Linux console.
>
> I can confirm. Branden, is this a bug? I think the Linux console
> should print the URI as in the old days.
I can't reproduce that behavior with groff 1.23.0 but I can with Git.
And that's due to a deliberate change.
NEWS[1]:
* Hyperlink support is now enabled by default on PDF and terminal
devices for an (man) and doc (mdoc) documents. Instructions and
commented code for disabling it are in the "man.local" and
"mdoc.local" files.
For the moment, groff's hands are kind of tied. To get this feature to
Just Work(TM) with no user involvement requires three things to happen.
1. grotty(1) needs to become a terminfo application. Lennart Jablonka
provided a nice proof of concept some 2 years ago,[2] and when I
started down the road of implementing it to fit the glorious visions
in my head, I discovered that I didn't know nearly enough about
terminfo or curses. I tried learning the system from its man pages
and books by Dan Gookin and Berny Goodheart. As you might guess, I
swiftly found myself dissatisfied with the currency and completeness
of (n)curses documentation, so I got involved with the project and
have submitted literally hundreds of patches to it, nearly all of
which Thomas Dickey has kindly accepted.
2. Once we have a tricked-out, terminfo-aware grotty, we meet another
problem:
There is no terminfo capability advertising OSC 8 hyperlink support.
That means that a terminal (emulator) can't _tell_ a terminfo
application whether it has the feature, and so grotty doesn't know
whether it should be emitting OSC 8 escape sequences or not.
I'll attach the bleeding edge version of the ncurses user_caps.5
page (which I recently revised to be more intelligible to knuckle
draggers like myself), because it provides essential background.
The community will need to select a (probably termcap-compatible,
and thus two-character) name for this extension capability. I
observe that "o8" is not taken. But neither are "ur" or "UR".
$ cd ncurses-snapshots
$ git grep -Eiw '(o8|ur)' include/Caps* || echo NO MATCH
NO MATCH
3. Even once we have a grotty(1) that knows whether the terminal device
can render hyperlinks or not, we have yet another problem:
Whether the output device supports hyperlinks has consequences on
the formatted content of the rendered document, because we don't
splat the URL into the output as formatted text unless (a) the
document author directs us to (by supplying empty link text for
groff man(7)'s `UR` and `UE` macros, for example), or (b) we have no
choice because we know the output device doesn't support hyperlinks.
Whether you format a URL or not can have significant consequences on
your page layout; they can change where your output lines break or
even change where the page breaks occur. Translation of device-
independent output to device-specific output is too late to be
making decisions that require re-laying out the page.
Thus, the formatted content of a document can depend on
properties of the output device in a new way, one more, uh,
"exciting" than the ~1972 binary choice of "is this an nroff device
[~terminal/typewriter]?" or "is this a troff device [typesetter]?",
or the ~1980 parameter "what is the name of the output device?"
(such as [Kernighan troff] "post", "l202", or [groff] "ps", "pdf",
"utf8").
The good news is that we already have a place for this information,
and the formatter already reads it when it initializes: the "DESC"
file. What I want is a _general_ interface for communicating any
information _via the groff language_ from that "DESC" file to a
document. While I haven't tried to implement this yet, it _seems_
like it shouldn't be too hard; the "DESC" file is a text file, and
the parameters (or "directives") it expresses are either Boolean-
valued (they are present or not) or string-valued.[3] Booleans can
be mapped to interpolations of "0" or "1", and strings can be
interpolated as-is (probably with newline elision [or space-
substitution thereof] for the quirky "fonts" and "sizes"
directives).
There aren't many mechanisms of adding something to the groff
language, so I reckon we'd end up with a new escape sequence (maybe
`\T`?) and, possibly, a corresponding request because previous groff
maintainers seem to have pursued parity between these syntactical
features. (For example, Clark added `device` and `brp` requests
corresponding to `\X` and `\p`.)
(However, I can't see what purpose a request would serve here
besides interpolating the property value into the document, which
could already be achieved with the escape sequence.)
Thus, assuming we added a "can-hyperlink" directive to the "DESC"
file format specification, and a `\T` escape sequence for exposing
it (and other properties) to the formatter, we could do something
like this in "tmac/an.tmac"...
.nr U \T'can-hyperlink'
...still respecting an overriding user specification, of course.
This stuff is going to take a while to land, even assuming no one finds
a fatal flaw in it, thus my advice above regarding user/site
configuration of groff man(7)'s `U` register.
Regards,
Branden
[1]
https://cgit.git.savannah.gnu.org/cgit/groff.git/tree/NEWS?id=c903a95f0a63607e336e431bb5cd07d2be6c6273#n355
[2] https://savannah.gnu.org/bugs/?63583
[3] There's an exception: the "charset" directive, specified and used by
Kernighan troff, but groff has neglected it for 35 years (see
groff_font(5)), and I don't propose to stop. In groff, the
"charset" property is coupled to a font, not the device per se.
'\" t
.\"***************************************************************************
.\" Copyright 2018-2024,2025 Thomas E. Dickey *
.\" Copyright 2017 Free Software Foundation, Inc. *
.\" *
.\" Permission is hereby granted, free of charge, to any person obtaining a *
.\" copy of this software and associated documentation files (the *
.\" "Software"), to deal in the Software without restriction, including *
.\" without limitation the rights to use, copy, modify, merge, publish, *
.\" distribute, distribute with modifications, sublicense, and/or sell *
.\" copies of the Software, and to permit persons to whom the Software is *
.\" furnished to do so, subject to the following conditions: *
.\" *
.\" The above copyright notice and this permission notice shall be included *
.\" in all copies or substantial portions of the Software. *
.\" *
.\" THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS *
.\" OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF *
.\" MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. *
.\" IN NO EVENT SHALL THE ABOVE COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, *
.\" DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR *
.\" OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR *
.\" THE USE OR OTHER DEALINGS IN THE SOFTWARE. *
.\" *
.\" Except as contained in this notice, the name(s) of the above copyright *
.\" holders shall not be used in advertising or otherwise to promote the *
.\" sale, use or other dealings in this Software without prior written *
.\" authorization. *
.\"***************************************************************************
.\"
.\" $Id: user_caps.5,v 1.58 2025/08/16 19:36:29 tom Exp $
.TH user_caps 5 2025-08-16 "ncurses 6.5" "File formats"
.ie \n(.g \{\
.ds `` \(lq
.ds '' \(rq
.\}
.el \{\
.ie t .ds `` ``
.el .ds `` ""
.ie t .ds '' ''
.el .ds '' ""
.\}
.
.de bP
.ie n .IP \(bu 4
.el .IP \(bu 2
..
.SH NAME
user_caps \-
user-defined \%\fIterm\%info\fP capability format
.SH SYNOPSIS
.B infocmp \-x
.PP
.B tic \-x
.SH DESCRIPTION
.SS Background
Prior to
.I \%ncurses
5.0 (1999),
.I \%term\%info
databases used a
.I "fixed repertoire"
of terminal capabilities designed
for the SVr2 terminal database in 1984,
added to in stages through SVr4 (1989),
and standardized in X/Open Curses starting in 1995.
.\" That date is a surmise based on the capability list appearing in
.\" Issue 4, Version 2 (1996). That list is not in man page format in
.\" the standard, so lacks a "HISTORY" section. However, `tigetstr()`
.\" and `tputs()` are identified in the same document as new to Issue 4,
.\" so GBR conjectures that the list came in at the same time.
.\"
.\" TED: the list is reflected in term.h, seen in examples from AIX 3 and 4,
.\" HP-UX 9, OSF/1, Solaris 2.4, dating from 1992-1994 -- all before 1996.
.\" The AIX 4 file has copyright dates starting in 1984;
.\" the Solaris file cites 1988 (the others have no copyright comments).
.\" Those term.h files note in a comment that it is generated by a script with
.\" a data file, i.e.,
.\" term.h - this file is automatically made from caps and maketerm.ex.
.\" illumos-gate has related source, with a "caps" file having AT&T copyright
.\" for 1988, and UCB copyright for 1982, 1986, 1988. That 1982 is interesting
.\" (hinting that something may have been in the initial releases of System V)
.\" but the first release with tic appears to be SVr2 in 1984.
.PP
Most such additions to this fixed repertoire
suppelmented the tables of Boolean,
numeric,
and string capabilities.
Rather than changing the meaning of an existing capability,
a new name was added.
The
.I \%term\%info
database uses a binary format;
binary compatibility was ensured
by using a header
that counted the number of items in the tables
for each type of capability.
Because each
.I curses
vendor extended the standard capability lists in distinct ways,
a library could be programmed to recognize only compiled
.I \%term\%info
entries that it was prepared to interpret.
Standardization was incomplete.
.bP
X/Open Curses describes only the
.I source
format,
not its
.I binary
representation on disk.
.IP
Library developers rely upon SVr4 documentation
and reverse engineering of compiled
.I \%term\%info
files to match the binary format.
.bP
Lacking a standard for the binary format,
most implementations copy the SVr2 binary format,
which uses 16-bit signed integers,
and is limited to 4096-byte entries.
.IP
The SVr2 format cannot represent very large numeric capability values,
nor can it represent large numbers of key definitions,
as are required to distinguish multiple modifier keys used
in combination with a function key.
.bP
The tables of capability names differ between implementations.
.IP
Although they
.I may
provide all of the standard capability names,
each arranges its table entries differently
because some features were added as needed,
while others were added
\(em out of order \(em
for X/Open Curses conformance.
.IP
While
.IR ncurses 's
capability repertoire is closest to that of Solaris,
the set of capabilities supported by
each vendor's
.I \%term\%info
database differs
from the list published by X/Open Curses.
.I \%ncurses
can be configured
with tables that match the terminal databases
for AIX,
HP-UX,
or OSF/1,
rather than the default Solaris-like configuration.
.bP
In SVr4
.I curses
and
.IR \%ncurses ","
the terminal database is defined at compile time
by interpolating a text file
that lists the different terminal capabilities.
.IP
In principle,
the text file can be extended,
but doing so requires recompiling and reinstalling the library.
The text file used by
.I \%ncurses
for terminal capabilities includes details of extensions
to X/Open Curses
made by various systems.
For example,
.I \%ncurses
supports the following nonstandard capabilities in each configuration.
.RS 8
.TP 5
.B memory_lock
.RB ( meml )
lock memory above cursor
.TP 5
.B memory_unlock
.RB ( memu )
unlock memory
.TP 5
.B box_chars_1
.RB ( box1 )
box characters primary set
.RE
.IP
The memory lock/unlock capabilities were included because they were used
in the X11R6 terminal description for \fIxterm\fP(1).
.B \%tic
uses the
.B box1
capability to cope with terminal descriptions written for AIX.
.PP
During the 1990s,
some application developers were reluctant to use
.I \%term\%info
in spite of its performance
(and other)
advantages over
.IR termcap "."
.bP
The fixed repertoire prevented users from adding support
for terminal features unanticipated by X/Open Curses
(or required them to reuse existing capabilities as a workaround).
.bP
The limitation to 16-bit signed integers was also mentioned.
Because
.I termcap
stores everything as a string,
it could represent larger numbers.
.PP
Although
.IR termcap 's
extensibility was rarely used
\(em the claimant was never an implementor
who had actually exercised it \(em
the criticism had a point.
.I \%ncurses
5.0 provided a way to detect nonstandard capabilities,
to determine their type,
and to optionally store and retrieve them
in a way that did not interfere with other applications.
.I ncurses
terms these
.I "user-defined capabilities"
because no modifications
to the standard capability list are needed.
.PP
The
.I \%ncurses
utilities
.B \%tic
and
.B \%infocmp
have a command-line option \*(``\-x\*(''
to control whether the nonstandard capabilities
are stored or retrieved.
.I \%ncurses
provides \fBuse_extended_names\fP(3NCURSES) to programs for the same purpose.
.PP
When compiling a terminal database, if \*(``\-x\*('' is used,
.B \%tic
stores a user-defined capability
if the capability name is not standard.
.PP
Because
.I \%ncurses
provides a
.I termcap
library interface,
these user-defined capabilities may be visible to
.I termcap
applications.
.bP
The
.I termcap
interface
(like all implementations of
.IR termcap )
restricts capability names to two characters.
.IP
When the capability is simple enough for use in a
.I termcap
application,
it is provided as a two-character name.
.bP
Other user-defined capabilities employ features not usable in
.IR termcap ","
such as parameterized strings that use more than two parameters
or require more powerful expressions than
.I termcap
supports.
Such capabilities should,
in the
.I \%term\%info
database,
have names at least three characters in length.
.bP
Some terminals can send distinct strings for special keys (cursor-,
keypad- or function-keys) depending on modifier keys (shift, control, etc.).
While
.I \%term\%info
and
.I termcap
define a set of sixty function key names,
to which a series of keys can be assigned,
that is insufficient for more than a dozen keys multiplied by more than
a couple of modifier combinations.
The
.I \%ncurses
database uses a convention based on \fIxterm\fP(1)
to provide extended special-key names.
.IP
Fitting that into
.IR termcap 's
limitation of 2-character names
would be pointless.
These extended keys are available only with
.IR \%term\%info "."
.SS "Recognized Capabilities"
The
.I \%ncurses
library employs user-definable capabilities.
While the
.I \%term\%info
database may have other extensions,
.I \%ncurses
makes explicit checks for the following.
.RS 3
.TP 3
.B AX
(Boolean)
asserts that the terminal interprets SGR 39 and SGR 49
by resetting the foreground and background colors,
respectively,
to the default.
.IP
\fBscreen\fP(1)
recognizes this capability as well.
.TP 3
.B E3
(string)
tells an application how to clear the terminal's scrollback buffer.
When present,
the \fBclear\fP(1) program sends this before clearing the terminal.
.IP
The command
.RB \*(`` "tput clear" \*(''
does the same thing.
.TP 3
.B NQ
(Boolean)
suppresses a consistency check in
.B \%tic
for the
.I \%ncurses
string capabilities
.B user6
.RB ( u6 )
through
.B user9
.RB ( u9 ),
which tell an application how to query the terminal's cursor position
and its device attributes.
.TP 3
.B RGB
(Boolean,
numeric,
or
string)
asserts that the
.B \%set_a_foreground
.RB ( setaf )
and
.B \%set_a_background
.RB ( setab )
capabilities employ
.IR "direct colors" ","
using an RGB (red/green/blue) convention.
This capability allows \fBcolor_content\fP(3NCURSES)
to return appropriate values
without requiring the application
to initialize colors using \fBinit_color\fP(3NCURSES).
.IP
The capability type determines the values
.I \%ncurses
sees.
.RS 3
.TP 3
Boolean
implies that the number of bits for red,
green,
and blue are the same.
Starting with the value of the capability
.B max_colors
.RB \%( colors ;
.I termcap:
.BR co ),
.I \%ncurses
adds two,
divides the sum by three,
and assigns the result to red,
green,
and blue,
in that order.
.IP
If the number of bits needed for the number of colors
is not a multiple of three,
the blue (and green) color channels lose in comparison to red.
.TP 3
numeric
tells
.I \%ncurses
what result to add to red,
green,
and blue.
If
.I \%ncurses
runs out of bits,
blue (and green) lose just as in the Boolean case.
.TP 3
string
specify the quantity of bits used for
red,
green,
and blue color channels
as a slash-separated list of decimal integers.
.RE
.IP
Because there are several RGB encodings in use,
applications that make assumptions
about the number of bits per color channel
are unlikely to work reliably.
As a trivial case,
one could define
.B RGB#1
to represent the standard eight ANSI\ X3.64/ECMA-48/ISO\ 6429 colors
using one bit per color channel.
.TP 3
.B U8
(numeric)
asserts whether
.I \%ncurses
must use Unicode values for line-drawing characters,
and that it should ignore the alternate character set (ACS) capabilities
when the locale uses UTF-8 encoding.
See the discussion of
.B \%NCURSES_NO_UTF8_ACS
in section \*(``ENVIRONMENT\*('' of \fB\%ncurses\fP(3NCURSES).
.IP
Set this capability to a nonzero value to enable it.
.TP 3
.B XM
(string)
override
.IR \%ncurses 's
built-in string that directs \fIxterm\fP(1)
to enable or disable mouse mode.
.IP
.I \%ncurses
sends a character sequence to the terminal to initialize mouse mode,
and when the user clicks the mouse buttons or
(in certain modes)
moves the mouse,
handles the characters sent back by the terminal
to tell the application what was done with the mouse.
.IP
The mouse protocol is enabled
when the
.I \fImask\fP
argument to the \fBmousemask\fP(3NCURSES) function is nonzero.
By default,
.I \%ncurses
handles the responses for the X11
.I xterm
mouse protocol.
It also knows about the SGR\ 1006
.I xterm
mouse protocol,
but must to be told to look for it specifically.
.I \%ncurses
is not be able to guess which of the two modes is used,
because the responses are enough alike that only confusion would result.
.IP
The
.B XM
capability has a single numeric parameter.
If nonzero,
the mouse protocol should be enabled.
If zero,
the mouse protocol should be disabled.
.I \%ncurses
inspects this capability if it is present,
to see whether the 1006 protocol is used.
If so,
it expects the responses to use the SGR\ 1006
.I xterm
mouse protocol.
.IP
The
.I xterm
mouse protocol is used by other terminal emulators.
The terminal database uses building blocks for the various
.I xterm
mouse protocols usable in customized terminal descriptions.
.IP
The terminal database building blocks for this mouse feature
also have an experimental capability,
.BR xm ","
that describes the mouse response.
No known interpreter uses this information,
which could make mouse support completely data-driven.
.IP
.B xm
shows the format of the mouse responses.
In this experimental capability,
the parameters are as follows.
.RS 5
.TP 5
.I p1
y-ordinate
.TP 5
.I p2
x-ordinate
.TP 5
.I p3
button
.TP 5
.I p4
state, e.g., pressed or released
.TP 5
.I p5
y-ordinate starting region
.TP 5
.I p6
x-ordinate starting region
.TP 5
.I p7
y-ordinate ending region
.TP 5
.I p8
x-ordinate ending region
.RE
.IP
Here are examples from the terminal database for the most commonly used
.I xterm
mouse protocols.
.IP
.EX
.nf
xterm+x11mouse|X11 xterm mouse protocol,
kmous=\eE[M, XM=\eE[?1000%?%p1%{1}%=%th%el%;,
xm=\eE[M
%?%p4%t%p3%e%{3}%;%'\ '%+%c
%p2%'!'%+%c
%p1%'!'%+%c,
.IP
xterm+sm+1006|xterm SGR-mouse,
kmous=\eE[<, XM=\eE[?1006;1000%?%p1%{1}%=%th%el%;,
xm=\eE[<%i%p3%d;
%p1%d;
%p2%d;
%?%p4%tM%em%;,
.fi
.EE
.
.SS "Extended Key Definitions"
Several terminals provide the ability to send distinct strings for
combinations of modified special keys.
There is no standard for what those keys can send.
.PP
Since 1999, \fIxterm\fP(1) has supported
\fIshift\fP, \fIcontrol\fP, \fIalt\fP, and \fImeta\fP modifiers which produce
distinct special-key strings.
In a terminal description,
\fI\%ncurses\fP has no special knowledge of the modifiers used.
Applications can use the \fInaming convention\fP established for
\fIxterm\fP
to find these special keys in the terminal description.
.PP
Starting with the
.I curses
convention that capability codes describing the input generated by a
terminal's key caps begin with \*(``k\*('',
and that shifted special keys use uppercase letters in their names,
.IR \%ncurses 's
terminal database defines the following names and codes to which a
suffix is added.
.PP
.RS 5
.TS
Lb Lb
Lb Lx.
Code Description
_
kDC shifted kdch1 (delete character)
.\" kDC is a standard capability; see X/Open Curses Issue 7, p. 345.
kDN shifted kcud1 (cursor down)
kEND shifted kend (end)
kHOM shifted khome (home)
kLFT shifted kcub1 (cursor back)
kNXT shifted knext (next)
kPRV shifted kprev (previous)
kRIT shifted kcuf1 (cursor forward)
kUP shifted kcuu1 (cursor up)
.TE
.RE
.PP
Keycap nomenclature on the Unix systems for which
.I curses
was developed differs from today's ubiquitous descendants of the IBM
PC/AT keyboard layout.
In the foregoing,
interpret \*(``backward\*('' as \*(``left\*('',
\*(``forward\*('' as \*(``right\*('',
\*(``next\*('' as \*(``page down\*('',
and
\*(``prev(ious)\*('' as \*(``page up\*(''.
.PP
These are the suffixes used to denote the modifiers:
.PP
.RS 5
.TS
tab(/) ;
l l .
\fBValue\fP/\fBDescription\fP
_
2/Shift
3/Alt
4/Shift + Alt
5/Control
6/Shift + Control
7/Alt + Control
8/Shift + Alt + Control
9/Meta
10/Meta + Shift
11/Meta + Alt
12/Meta + Alt + Shift
13/Meta + Ctrl
14/Meta + Ctrl + Shift
15/Meta + Ctrl + Alt
16/Meta + Ctrl + Alt + Shift
.TE
.RE
.PP
.I ncurses
defines no capabilities for modified F-keys;
terminal descriptions can refer to
.I names
that
.I \%ncurses
allocates at runtime to
.IR "key codes" "."
To use these keys in an \fI\%ncurses\fP program,
an application could do this:
.bP
using a list of extended key \fInames\fP,
ask \fBtigetstr\fP(3NCURSES) for their values, and
.bP
given the list of values,
ask \fBkey_defined\fP(3NCURSES) for the \fIkey-code\fP which
would be returned for those keys by \fBwgetch\fP(3NCURSES).
.\"
.SH PORTABILITY
The \*(``\-x\*('' extension feature of
.B \%tic
and
.B \%infocmp
has been adopted in NetBSD
.IR curses "."
That implementation stores user-defined capabilities,
but makes no use of these capabilities itself.
.\"
.SH AUTHORS
Thomas E. Dickey
.br
beginning with
.I \%ncurses
5.0 (1999)
.\"
.SH SEE ALSO
\fB\%infocmp\fP(1),
\fB\%tic\fP(1)
.PP
In the source form of the terminal database,
.IR \%terminfo.src ","
the section \*(``NCURSES USER-DEFINABLE CAPABILITIES\*(''.
summarizes commonly-used user-defined capabilities
employed in the terminal descriptions.
Some of those features are mentioned in \fB\%screen\fP(1) or
\fBtmux\fP(1).
.PP
.I "XTerm Control Sequences"
provides further information on the \fI\%xterm\fP(1) features
that are used in these extended capabilities.
signature.asc
Description: PGP signature
