path: root/README.FAQ

RXVT-UNICODE/URXVT FREQUENTLY ASKED QUESTIONS
  Meta, Features & Commandline Issues
   My question isn't answered here, can I ask a human?
    Before sending me mail, you could go to IRC: "irc.freenode.net", channel
    "#rxvt-unicode" has some rxvt-unicode enthusiasts that might be
    interested in learning about new and exciting problems (but not FAQs :).

   I use Gentoo, and I have a problem...
    There are two big problems with Gentoo Linux: first, most if not all
    Gentoo systems are completely broken (missing or mismatched header
    files, broken compiler etc. are just the tip of the iceberg); secondly,
    it should be called Gentoo GNU/Linux.

    For these reasons, it is impossible to support rxvt-unicode on Gentoo.
    Problems appearing on Gentoo systems will usually simply be ignored
    unless they can be reproduced on non-Gentoo systems.

   Does it support tabs, can I have a tabbed rxvt-unicode?
    Beginning with version 7.3, there is a perl extension that implements a
    simple tabbed terminal. It is installed by default, so any of these
    should give you tabs:

       urxvt -pe tabbed

       URxvt.perl-ext-common: default,tabbed

    It will also work fine with tabbing functionality of many window
    managers or similar tabbing programs, and its embedding-features allow
    it to be embedded into other programs, as witnessed by doc/rxvt-tabbed
    or the upcoming "Gtk2::URxvt" perl module, which features a tabbed urxvt
    (murxvt) terminal as an example embedding application.

   How do I know which rxvt-unicode version I'm using?
    The version number is displayed with the usage (-h). Also the escape
    sequence "ESC [ 8 n" sets the window title to the version number. When
    using the urxvtc client, the version displayed is that of the daemon.

   Rxvt-unicode uses gobs of memory, how can I reduce that?
    Rxvt-unicode tries to obey the rule of not charging you for something
    you don't use. One thing you should try is to configure out all settings
    that you don't need, for example, Xft support is a resource hog by
    design, when used. Compiling it out ensures that no Xft font will be
    loaded accidentally when rxvt-unicode tries to find a font for your
    characters.

    Also, many people (me included) like large windows and even larger
    scrollback buffers: Without "--enable-unicode3", rxvt-unicode will use 6
    bytes per screen cell. For a 160x?? window this amounts to almost a
    kilobyte per line. A scrollback buffer of 10000 lines will then (if
    full) use 10 Megabytes of memory. With "--enable-unicode3" it gets
    worse, as rxvt-unicode then uses 8 bytes per screen cell.

   How can I start urxvtd in a race-free way?
    Try "urxvtd -f -o", which tells urxvtd to open the display, create the
    listening socket and then fork.

   How can I start urxvtd automatically when I run urxvtc?
    If you want to start urxvtd automatically whenever you run urxvtc and
    the daemon isn't running yet, use this script:

       #!/bin/sh
       urxvtc "$@"
       if [ $? -eq 2 ]; then
          urxvtd -q -o -f
          urxvtc "$@"
       fi

    This tries to create a new terminal, and if fails with exit status 2,
    meaning it couldn't connect to the daemon, it will start the daemon and
    re-run the command. Subsequent invocations of the script will re-use the
    existing daemon.

   How do I distinguish whether I'm running rxvt-unicode or a regular
xterm? I need this to decide about setting colours etc.
    The original rxvt and rxvt-unicode always export the variable
    "COLORTERM", so you can check and see if that is set. Note that several
    programs, JED, slrn, Midnight Commander automatically check this
    variable to decide whether or not to use colour.

   How do I set the correct, full IP address for the DISPLAY variable?
    If you've compiled rxvt-unicode with DISPLAY_IS_IP and have enabled
    insecure mode then it is possible to use the following shell script
    snippets to correctly set the display. If your version of rxvt-unicode
    wasn't also compiled with ESCZ_ANSWER (as assumed in these snippets)
    then the COLORTERM variable can be used to distinguish rxvt-unicode from
    a regular xterm.

    Courtesy of Chuck Blake <cblake@BBN.COM> with the following shell script
    snippets:

       # Bourne/Korn/POSIX family of shells:
       [ ${TERM:-foo} = foo ] && TERM=xterm # assume an xterm if we don't know
       if [ ${TERM:-foo} = xterm ]; then
          stty -icanon -echo min 0 time 15 # see if enhanced rxvt or not
          printf "\eZ"
          read term_id
          stty icanon echo
          if [ ""${term_id} = '^[[?1;2C' -a ${DISPLAY:-foo} = foo ]; then
             printf '\e[7n'        # query the rxvt we are in for the DISPLAY string
             read DISPLAY          # set it in our local shell
          fi
       fi

   How do I compile the manual pages on my own?
    You need to have a recent version of perl installed as /usr/bin/perl,
    one that comes with pod2man, pod2text and pod2xhtml (from Pod::Xhtml).
    Then go to the doc subdirectory and enter "make alldoc".

   Isn't rxvt-unicode supposed to be small? Don't all those features bloat?
    I often get asked about this, and I think, no, they didn't cause extra
    bloat. If you compare a minimal rxvt and a minimal urxvt, you can see
    that the urxvt binary is larger (due to some encoding tables always
    being compiled in), but it actually uses less memory (RSS) after
    startup. Even with "--disable-everything", this comparison is a bit
    unfair, as many features unique to urxvt (locale, encoding conversion,
    iso14755 etc.) are already in use in this mode.

        text    data     bss     drs     rss filename
       98398    1664      24   15695    1824 rxvt --disable-everything
      188985    9048   66616   18222    1788 urxvt --disable-everything

    When you "--enable-everything" (which *is* unfair, as this involves xft
    and full locale/XIM support which are quite bloaty inside libX11 and my
    libc), the two diverge, but not unreasonably so.

        text    data     bss     drs     rss filename
      163431    2152      24   20123    2060 rxvt --enable-everything
     1035683   49680   66648   29096    3680 urxvt --enable-everything

    The very large size of the text section is explained by the east-asian
    encoding tables, which, if unused, take up disk space but nothing else
    and can be compiled out unless you rely on X11 core fonts that use those
    encodings. The BSS size comes from the 64k emergency buffer that my c++
    compiler allocates (but of course doesn't use unless you are out of
    memory). Also, using an xft font instead of a core font immediately adds
    a few megabytes of RSS. Xft indeed is responsible for a lot of RSS even
    when not used.

    Of course, due to every character using two or four bytes instead of
    one, a large scrollback buffer will ultimately make rxvt-unicode use
    more memory.

    Compared to e.g. Eterm (5112k), aterm (3132k) and xterm (4680k), this
    still fares rather well. And compared to some monsters like
    gnome-terminal (21152k + extra 4204k in separate processes) or konsole
    (22200k + extra 43180k in daemons that stay around after exit, plus half
    a minute of startup time, including the hundreds of warnings it spits
    out), it fares extremely well *g*.

   Why C++, isn't that unportable/bloated/uncool?
    Is this a question? :) It comes up very often. The simple answer is: I
    had to write it, and C++ allowed me to write and maintain it in a
    fraction of the time and effort (which is a scarce resource for me). Put
    even shorter: It simply wouldn't exist without C++.

    My personal stance on this is that C++ is less portable than C, but in
    the case of rxvt-unicode this hardly matters, as its portability limits
    are defined by things like X11, pseudo terminals, locale support and
    unix domain sockets, which are all less portable than C++ itself.

    Regarding the bloat, see the above question: It's easy to write programs
    in C that use gobs of memory, and certainly possible to write programs
    in C++ that don't. C++ also often comes with large libraries, but this
    is not necessarily the case with GCC. Here is what rxvt links against on
    my system with a minimal config:

       libX11.so.6 => /usr/X11R6/lib/libX11.so.6 (0x00002aaaaabc3000)
       libc.so.6 => /lib/libc.so.6 (0x00002aaaaadde000)
       libdl.so.2 => /lib/libdl.so.2 (0x00002aaaab01d000)
       /lib64/ld-linux-x86-64.so.2 (0x00002aaaaaaab000)

    And here is rxvt-unicode:

       libX11.so.6 => /usr/X11R6/lib/libX11.so.6 (0x00002aaaaabc3000)
       libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x00002aaaaada2000)
       libc.so.6 => /lib/libc.so.6 (0x00002aaaaaeb0000)
       libdl.so.2 => /lib/libdl.so.2 (0x00002aaaab0ee000)
       /lib64/ld-linux-x86-64.so.2 (0x00002aaaaaaab000)

    No large bloated libraries (of course, none were linked in statically),
    except maybe libX11 :)

  Rendering, Font & Look and Feel Issues
   I can't get transparency working, what am I doing wrong?
    First of all, transparency isn't officially supported in rxvt-unicode,
    so you are mostly on your own. Do not bug the author about it (but you
    may bug everybody else). Also, if you can't get it working consider it a
    rite of passage: ... and you failed.

    Here are four ways to get transparency. Do read the manpage and option
    descriptions for the programs mentioned and rxvt-unicode. Really, do it!

    1. Use transparent mode:

       Esetroot wallpaper.jpg
       urxvt -tr -tint red -sh 40

    That works. If you think it doesn't, you lack transparency and tinting
    support, or you are unable to read. This method requires that the
    background-setting program sets the _XROOTPMAP_ID or ESETROOT_PMAP_ID
    property. Compatible programs are Esetroot, hsetroot and feh.

    2. Use a simple pixmap and emulate pseudo-transparency. This enables you
    to use effects other than tinting and shading: Just shade/tint/whatever
    your picture with gimp or any other tool:

       convert wallpaper.jpg -blur 20x20 -modulate 30 background.jpg
       urxvt -pixmap "background.jpg;:root"

    That works. If you think it doesn't, you lack GDK-PixBuf support, or you
    are unable to read.

    3. Use an ARGB visual:

       urxvt -depth 32 -fg grey90 -bg rgba:0000/0000/4444/cccc

    This requires XFT support, and the support of your X-server. If that
    doesn't work for you, blame Xorg and Keith Packard. ARGB visuals aren't
    there yet, no matter what they claim. Rxvt-Unicode contains the
    necessary bugfixes and workarounds for Xft and Xlib to make it work, but
    that doesn't mean that your WM has the required kludges in place.

    4. Use xcompmgr and let it do the job:

      xprop -frame -f _NET_WM_WINDOW_OPACITY 32c \
            -set _NET_WM_WINDOW_OPACITY 0xc0000000

    Then click on a window you want to make transparent. Replace 0xc0000000
    by other values to change the degree of opacity. If it doesn't work and
    your server crashes, you got to keep the pieces.

   Why does rxvt-unicode sometimes leave pixel droppings?
    Most fonts were not designed for terminal use, which means that
    character size varies a lot. A font that is otherwise fine for terminal
    use might contain some characters that are simply too wide. Rxvt-unicode
    will avoid these characters. For characters that are just "a bit" too
    wide a special "careful" rendering mode is used that redraws adjacent
    characters.

    All of this requires that fonts do not lie about character sizes,
    however: Xft fonts often draw glyphs larger than their acclaimed
    bounding box, and rxvt-unicode has no way of detecting this (the correct
    way is to ask for the character bounding box, which unfortunately is
    wrong in these cases).

    It's not clear (to me at least), whether this is a bug in Xft, freetype,
    or the respective font. If you encounter this problem you might try
    using the "-lsp" option to give the font more height. If that doesn't
    work, you might be forced to use a different font.

    All of this is not a problem when using X11 core fonts, as their
    bounding box data is correct.

   How can I keep rxvt-unicode from using reverse video so much?
    First of all, make sure you are running with the right terminal settings
    ("TERM=rxvt-unicode"), which will get rid of most of these effects. Then
    make sure you have specified colours for italic and bold, as otherwise
    rxvt-unicode might use reverse video to simulate the effect:

       URxvt.colorBD:  white
       URxvt.colorIT:  green

   Some programs assume totally weird colours (red instead of blue), how can I fix that?
    For some unexplainable reason, some rare programs assume a very weird
    colour palette when confronted with a terminal with more than the
    standard 8 colours (rxvt-unicode supports 88). The right fix is, of
    course, to fix these programs not to assume non-ISO colours without very
    good reasons.

    In the meantime, you can either edit your "rxvt-unicode" terminfo
    definition to only claim 8 colour support or use "TERM=rxvt", which will
    fix colours but keep you from using other rxvt-unicode features.

   Can I switch the fonts at runtime?
    Yes, using an escape sequence. Try something like this, which has the
    same effect as using the "-fn" switch, and takes effect immediately:

       printf '\33]50;%s\007' "9x15bold,xft:Kochi Gothic"

    This is useful if you e.g. work primarily with japanese (and prefer a
    japanese font), but you have to switch to chinese temporarily, where
    japanese fonts would only be in your way.

    You can think of this as a kind of manual ISO-2022 switching.

   Why do italic characters look as if clipped?
    Many fonts have difficulties with italic characters and hinting. For
    example, the otherwise very nicely hinted font "xft:Bitstream Vera Sans
    Mono" completely fails in its italic face. A workaround might be to
    enable freetype autohinting, i.e. like this:

       URxvt.italicFont:        xft:Bitstream Vera Sans Mono:italic:autohint=true
       URxvt.boldItalicFont:    xft:Bitstream Vera Sans Mono:bold:italic:autohint=true

   Can I speed up Xft rendering somehow?
    Yes, the most obvious way to speed it up is to avoid Xft entirely, as it
    is simply slow. If you still want Xft fonts you might try to disable
    antialiasing (by appending ":antialias=false"), which saves lots of
    memory and also speeds up rendering considerably.

   Rxvt-unicode doesn't seem to anti-alias its fonts, what is wrong?
    Rxvt-unicode will use whatever you specify as a font. If it needs to
    fall back to its default font search list it will prefer X11 core fonts,
    because they are small and fast, and then use Xft fonts. It has
    antialiasing disabled for most of them, because the author thinks they
    look best that way.

    If you want antialiasing, you have to specify the fonts manually.

   What's with this bold/blink stuff?
    If no bold colour is set via "colorBD:", bold will invert text using the
    standard foreground colour.

    For the standard background colour, blinking will actually make the text
    blink when compiled with "--enable-text-blink". Without
    "--enable-text-blink", the blink attribute will be ignored.

    On ANSI colours, bold/blink attributes are used to set high-intensity
    foreground/background colours.

    color0-7 are the low-intensity colours.

    color8-15 are the corresponding high-intensity colours.

   I don't like the screen colours.  How do I change them?
    You can change the screen colours at run-time using ~/.Xdefaults
    resources (or as long-options).

    Here are values that are supposed to resemble a VGA screen, including
    the murky brown that passes for low-intensity yellow:

       URxvt.color0:   #000000
       URxvt.color1:   #A80000
       URxvt.color2:   #00A800
       URxvt.color3:   #A8A800
       URxvt.color4:   #0000A8
       URxvt.color5:   #A800A8
       URxvt.color6:   #00A8A8
       URxvt.color7:   #A8A8A8

       URxvt.color8:   #000054
       URxvt.color9:   #FF0054
       URxvt.color10:  #00FF54
       URxvt.color11:  #FFFF54
       URxvt.color12:  #0000FF
       URxvt.color13:  #FF00FF
       URxvt.color14:  #00FFFF
       URxvt.color15:  #FFFFFF

    And here is a more complete set of non-standard colours.

       URxvt.cursorColor:  #dc74d1
       URxvt.pointerColor: #dc74d1
       URxvt.background:   #0e0e0e
       URxvt.foreground:   #4ad5e1
       URxvt.color0:       #000000
       URxvt.color8:       #8b8f93
       URxvt.color1:       #dc74d1
       URxvt.color9:       #dc74d1
       URxvt.color2:       #0eb8c7
       URxvt.color10:      #0eb8c7
       URxvt.color3:       #dfe37e
       URxvt.color11:      #dfe37e
       URxvt.color5:       #9e88f0
       URxvt.color13:      #9e88f0
       URxvt.color6:       #73f7ff
       URxvt.color14:      #73f7ff
       URxvt.color7:       #e1dddd
       URxvt.color15:      #e1dddd

    They have been described (not by me) as "pretty girly".

   Why do some characters look so much different than others?
    See next entry.

   How does rxvt-unicode choose fonts?
    Most fonts do not contain the full range of Unicode, which is fine.
    Chances are that the font you (or the admin/package maintainer of your
    system/os) have specified does not cover all the characters you want to
    display.

    rxvt-unicode makes a best-effort try at finding a replacement font.
    Often the result is fine, but sometimes the chosen font looks
    bad/ugly/wrong. Some fonts have totally strange characters that don't
    resemble the correct glyph at all, and rxvt-unicode lacks the artificial
    intelligence to detect that a specific glyph is wrong: it has to believe
    the font that the characters it claims to contain indeed look correct.

    In that case, select a font of your taste and add it to the font list,
    e.g.:

       urxvt -fn basefont,font2,font3...

    When rxvt-unicode sees a character, it will first look at the base font.
    If the base font does not contain the character, it will go to the next
    font, and so on. Specifying your own fonts will also speed up this
    search and use less resources within rxvt-unicode and the X-server.

    The only limitation is that none of the fonts may be larger than the
    base font, as the base font defines the terminal character cell size,
    which must be the same due to the way terminals work.

   Why do some chinese characters look so different than others?
    This is because there is a difference between script and language --
    rxvt-unicode does not know which language the text that is output is, as
    it only knows the unicode character codes. If rxvt-unicode first sees a
    japanese/chinese character, it might choose a japanese font for display.
    Subsequent japanese characters will use that font. Now, many chinese
    characters aren't represented in japanese fonts, so when the first
    non-japanese character comes up, rxvt-unicode will look for a chinese
    font -- unfortunately at this point, it will still use the japanese font
    for chinese characters that are also in the japanese font.

    The workaround is easy: just tag a chinese font at the end of your font
    list (see the previous question). The key is to view the font list as a
    preference list: If you expect more japanese, list a japanese font
    first. If you expect more chinese, put a chinese font first.

    In the future it might be possible to switch language preferences at
    runtime (the internal data structure has no problem with using different
    fonts for the same character at the same time, but no interface for this
    has been designed yet).

    Until then, you might get away with switching fonts at runtime (see "Can
    I switch the fonts at runtime?" later in this document).

   How can I make mplayer display video correctly?
    We are working on it, in the meantime, as a workaround, use something
    like:

       urxvt -b 600 -geometry 20x1 -e sh -c 'mplayer -wid $WINDOWID file...'

   Why is the cursor now blinking in emacs/vi/...?
    This is likely caused by your editor/program's use of the "cvvis"
    terminfo capability. Emacs uses it by default, as well as some versions
    of vi and possibly other programs.

    In emacs, you can switch that off by adding this to your ".emacs" file:

       (setq visible-cursor nil)

    For other programs, if they do not have an option, your have to remove
    the "cvvis" capability from the terminfo description.

    When urxvt first added the blinking cursor option, it didn't add a
    "cvvis" capability, which served no purpose before. Version 9.21
    introduced "cvvis" (and the ability to control blinking independent of
    cursor shape) for compatibility with other terminals, which
    traditionally use a blinking cursor for "cvvis". This also reflects the
    intent of programs such as emacs, who expect "cvvis" to enable a
    blinking cursor.

  Keyboard, Mouse & User Interaction
   The new selection selects pieces that are too big, how can I select single words?
    If you want to select e.g. alphanumeric words, you can use the following
    setting:

       URxvt.selection.pattern-0: ([[:word:]]+)

    If you click more than twice, the selection will be extended more and
    more.

    To get a selection that is very similar to the old code, try this
    pattern:

       URxvt.selection.pattern-0: ([^"&'()*,;<=>?@[\\\\]^`{|})]+)

    Please also note that the *LeftClick Shift-LeftClick* combination also
    selects words like the old code.

   I don't like the new selection/popups/hotkeys/perl, how do I change/disable it?
    You can disable the perl extension completely by setting the
    perl-ext-common resource to the empty string, which also keeps
    rxvt-unicode from initialising perl, saving memory.

    If you only want to disable specific features, you first have to
    identify which perl extension is responsible. For this, read the section
    PREPACKAGED EXTENSIONS in the urxvtperl(3) manpage. For example, to
    disable the selection-popup and option-popup, specify this
    perl-ext-common resource:

       URxvt.perl-ext-common: default,-selection-popup,-option-popup

    This will keep the default extensions, but disable the two popup
    extensions. Some extensions can also be configured, for example,
    scrollback search mode is triggered by M-s. You can move it to any other
    combination by adding a keysym resource that binds the desired
    combination to the "start" action of "searchable-scrollback" and another
    one that binds M-s to the "builtin:" action:

       URxvt.keysym.CM-s: searchable-scrollback:start
       URxvt.keysym.M-s: builtin:

   The cursor moves when selecting text in the current input line, how do I switch this off?
    See next entry.

   During rlogin/ssh/telnet/etc. sessions, clicking near the cursor outputs strange escape sequences, how do I fix this?
    These are caused by the "readline" perl extension. Under normal
    circumstances, it will move your cursor around when you click into the
    line that contains it. It tries hard not to do this at the wrong moment,
    but when running a program that doesn't parse cursor movements or in
    some cases during rlogin sessions, it fails to detect this properly.

    You can permanently switch this feature off by disabling the "readline"
    extension:

       URxvt.perl-ext-common: default,-readline

   My numeric keypad acts weird and generates differing output?
    Some Debian GNU/Linux users seem to have this problem, although no
    specific details were reported so far. It is possible that this is
    caused by the wrong "TERM" setting, although the details of whether and
    how this can happen are unknown, as "TERM=rxvt" should offer a
    compatible keymap. See the answer to the previous question, and please
    report if that helped.

   My Compose (Multi_key) key is no longer working.
    The most common causes for this are that either your locale is not set
    correctly, or you specified a preeditType that is not supported by your
    input method. For example, if you specified OverTheSpot and your input
    method (e.g. the default input method handling Compose keys) does not
    support this (for instance because it is not visual), then rx