16 months agochore: change alpha master
Artjom Löbsack [Tue, 11 May 2021 22:35:33 +0000 (03:35 +0500)]
chore: change alpha

17 months agorefactor: change alpha
Artjom Löbsack [Fri, 30 Apr 2021 23:12:32 +0000 (04:12 +0500)]
refactor: change alpha

17 months agorevert(fonts): remove fallback nerd font
Artjom Löbsack [Sun, 11 Apr 2021 22:41:19 +0000 (03:41 +0500)]
revert(fonts): remove fallback nerd font

17 months agofeat(fonts): add fallback nerd font
Artjom Löbsack [Sun, 11 Apr 2021 22:30:06 +0000 (03:30 +0500)]
feat(fonts): add fallback nerd font

18 months agorevert: remove 256color postfix to term name to show colors
Artjom Löbsack [Sat, 3 Apr 2021 17:13:06 +0000 (22:13 +0500)]
revert: remove 256color postfix to term name to show colors

18 months agofix: add 256color postfix to term name to show colors
Artjom Löbsack [Sat, 3 Apr 2021 17:02:31 +0000 (22:02 +0500)]
fix: add 256color postfix to term name to show colors

18 months agorevert: return borderpx
Artjom Löbsack [Sat, 3 Apr 2021 16:50:06 +0000 (21:50 +0500)]
revert: return borderpx

18 months agofeat: add bright colors
Artjom Löbsack [Sat, 3 Apr 2021 16:47:04 +0000 (21:47 +0500)]
feat: add bright colors

18 months agorefactor: remove border
Artjom Löbsack [Sat, 3 Apr 2021 16:35:25 +0000 (21:35 +0500)]
refactor: remove border

18 months agorefactor: change emoji size
Artjom Löbsack [Sat, 3 Apr 2021 16:32:00 +0000 (21:32 +0500)]
refactor: change emoji size

18 months agorefactor: change colors
Artjom Löbsack [Sat, 3 Apr 2021 02:29:06 +0000 (07:29 +0500)]
refactor: change colors

18 months agorefactor: change font size and alpha
Artjom Löbsack [Sat, 3 Apr 2021 00:38:13 +0000 (05:38 +0500)]
refactor: change font size and alpha

18 months agorefactor(fonts): change size
Artjom Löbsack [Fri, 2 Apr 2021 12:47:50 +0000 (17:47 +0500)]
refactor(fonts): change size

18 months agorefactor(fonts): use default font for emoji
Artjom Löbsack [Thu, 1 Apr 2021 14:22:10 +0000 (19:22 +0500)]
refactor(fonts): use default font for emoji

18 months agorefactor: remove alpha
Artjom Löbsack [Thu, 1 Apr 2021 14:06:32 +0000 (19:06 +0500)]
refactor: remove alpha

18 months agorefactor: change default font to monospace
Artjom Löbsack [Thu, 1 Apr 2021 14:05:40 +0000 (19:05 +0500)]
refactor: change default font to monospace

18 months agofeat: reduce emoji font size
Artjom Löbsack [Thu, 25 Mar 2021 18:34:30 +0000 (23:34 +0500)]
feat: reduce emoji font size

18 months agoMerge branch 'master' of
Artjom Löbsack [Thu, 25 Mar 2021 17:56:04 +0000 (22:56 +0500)]
Merge branch 'master' of

18 months agofix: correctly encode mouse buttons >= 8 in X10 and SGR mode
Hiltjo Posthuma [Fri, 19 Mar 2021 10:54:36 +0000 (11:54 +0100)]
fix: correctly encode mouse buttons >= 8 in X10 and SGR mode

These are typically mapped in X11 to the side-buttons (backward/forwards) on
the mouse. A comparison of the button numbers in SGR mode (first field):

st old:
0 1 2 64 65 66 67 68 69 70

st new (it is the same as xterm now):
0 1 2 64 65 66 67 128 129 130

A script to test and reproduce it, first argument is "h" (on) or "l" (off):

printf '\x1b[?1000%s\x1b[?1006%s' "$1" "$1"

for n in 1 2 3 4 5 6 7 8 9 10; do
printf 'button %d\n' "$n"
xdotool click "$n"
printf '\n\n'

21 months agorefactor: change alpha
Artjom Löbsack [Wed, 9 Dec 2020 19:21:48 +0000 (00:21 +0500)]
refactor: change alpha

21 months agofix(font): use fixed width
Artjom Löbsack [Sat, 5 Dec 2020 13:22:25 +0000 (18:22 +0500)]
fix(font): use fixed width

22 months agofeat: add backup bold font
Artjom Löbsack [Tue, 10 Nov 2020 18:39:42 +0000 (23:39 +0500)]
feat: add backup bold font

22 months agofix(fonts): use nerd fonts
Artjom Löbsack [Tue, 10 Nov 2020 14:49:42 +0000 (19:49 +0500)]
fix(fonts): use nerd fonts

22 months agochore: add config.h to gitignore
Artjom Löbsack [Tue, 10 Nov 2020 14:48:11 +0000 (19:48 +0500)]
chore: add config.h to gitignore

22 months agoMerge branch 'master' of
Artjom Löbsack [Tue, 3 Nov 2020 12:26:17 +0000 (17:26 +0500)]
Merge branch 'master' of

23 months agoremove unused variable from previous patch
Hiltjo Posthuma [Sun, 18 Oct 2020 09:18:03 +0000 (11:18 +0200)]
remove unused variable from previous patch

23 months agoST: Add WM_ICON_NAME property support
John Collis [Sun, 6 Sep 2020 05:53:41 +0000 (17:53 +1200)]
ST: Add WM_ICON_NAME property support

Also added _NET_WM_ICON_NAME.

23 months agofix: replace white color
Artjom Löbsack [Sat, 17 Oct 2020 07:03:11 +0000 (12:03 +0500)]
fix: replace white color

23 months agofix: patch ligatures with boxdraw
Artjom Löbsack [Sat, 17 Oct 2020 06:44:51 +0000 (11:44 +0500)]
fix: patch ligatures with boxdraw

23 months agofeat: patch ligatures
Artjom Löbsack [Sat, 17 Oct 2020 06:05:26 +0000 (11:05 +0500)]
feat: patch ligatures

23 months agorefactor: change some colors
Artjom Löbsack [Sat, 17 Oct 2020 05:47:19 +0000 (10:47 +0500)]
refactor: change some colors

23 months agofeat: patch boxdraw
Artjom Löbsack [Sat, 17 Oct 2020 05:40:48 +0000 (10:40 +0500)]
feat: patch boxdraw

23 months agofeat: patch openclipboard
Artjom Löbsack [Sat, 17 Oct 2020 05:15:31 +0000 (10:15 +0500)]
feat: patch openclipboard

23 months agofeat: patch anysize
Artjom Löbsack [Sat, 17 Oct 2020 05:07:57 +0000 (10:07 +0500)]
feat: patch anysize

23 months agorefactor: change shell path
Artjom Löbsack [Sat, 17 Oct 2020 05:03:51 +0000 (10:03 +0500)]
refactor: change shell path

23 months agofeat: patch newterm
Artjom Löbsack [Sat, 17 Oct 2020 04:42:07 +0000 (09:42 +0500)]
feat: patch newterm

23 months agofeat: patch copyurl
Artjom Löbsack [Sat, 17 Oct 2020 04:28:12 +0000 (09:28 +0500)]
feat: patch copyurl

23 months agorefactor: change bg opacity
Artjom Löbsack [Tue, 13 Oct 2020 22:59:21 +0000 (03:59 +0500)]
refactor: change bg opacity

23 months agorefactor: change bg colors
Artjom Löbsack [Tue, 13 Oct 2020 22:51:25 +0000 (03:51 +0500)]
refactor: change bg colors

23 months agofeat: patch scrollback
Artjom Löbsack [Tue, 13 Oct 2020 22:39:43 +0000 (03:39 +0500)]
feat: patch scrollback

23 months agofeat: patch alpha
Artjom Löbsack [Tue, 13 Oct 2020 22:24:50 +0000 (03:24 +0500)]
feat: patch alpha

23 months agorefactor: use config.def
Artjom Löbsack [Tue, 13 Oct 2020 21:56:46 +0000 (02:56 +0500)]
refactor: use config.def

23 months agorefactor: change default fontsize
Artjom Löbsack [Tue, 13 Oct 2020 19:54:20 +0000 (00:54 +0500)]
refactor: change default fontsize

2 years agochore: add .gitignore
Artjom Löbsack [Tue, 29 Sep 2020 16:52:58 +0000 (21:52 +0500)]
chore: add .gitignore

2 years agorefactor(config): edit font and theme
Artjom Löbsack [Tue, 29 Sep 2020 16:48:36 +0000 (21:48 +0500)]
refactor(config): edit font and theme

2 years agobump version to 0.8.4
Hiltjo Posthuma [Fri, 19 Jun 2020 09:27:17 +0000 (11:27 +0200)]
bump version to 0.8.4

2 years use PKG_CONFIG in commented OpenBSD section
Hiltjo Posthuma [Wed, 17 Jun 2020 21:47:00 +0000 (23:47 +0200)] use PKG_CONFIG in commented OpenBSD section

2 years agoLICENSE: bump years
Hiltjo Posthuma [Wed, 17 Jun 2020 21:44:34 +0000 (23:44 +0200)]
LICENSE: bump years

2 years agoremove sixel stub code
Hiltjo Posthuma [Wed, 17 Jun 2020 20:05:48 +0000 (22:05 +0200)]
remove sixel stub code

Remove stub code that was used for an experiment of adding sixel code to st
from the commit f7398434.

2 years agofix unicode glitch in DCS strings, patch by Tim Allen
Hiltjo Posthuma [Wed, 17 Jun 2020 19:35:39 +0000 (21:35 +0200)]
fix unicode glitch in DCS strings, patch by Tim Allen

Reported on the mailinglist:

I discovered recently that if an application running inside st tries to
send a DCS string, subsequent Unicode characters get messed up. For
example, consider the following test-case:

    printf '\303\277\033P\033\\\303\277'


  - \303\277 is the UTF-8 encoding of U+00FF LATIN SMALL LETTER Y WITH
    DIAERESIS (ÿ).
  - \033P is ESC P, the token that begins a DCS string.
  - \033\\ is ESC \, a token that ends a DCS string.
  - \303\277 is the same ÿ character again.

If I run the above command in a VTE-based terminal, or xterm, or
QTerminal, or pterm (PuTTY), I get the output:


...which is to say, the empty DCS string is ignored. However, if I run
that command inside st (as of commit 9ba7ecf), I get:


...where those last two characters are \303\277 interpreted as ISO8859-1
characters, instead of UTF-8.

I spent some time tracing through the state machines in st.c, and so far
as I can tell, this is how it works currently:

  - ESC P sets the "ESC_DCS" and "ESC_STR" flags, indicating that
    incoming bytes should be collected into the strescseq buffer, rather
    than being interpreted.
  - ESC \ sets the "ESC_STR_END" flag (when ESC is received), and then
    calls strhandle() (when \ is received) to interpret the collected
  - If the collected bytes begin with 'P' (i.e. if this was a DCS
    string) strhandle() sets the "ESC_DCS" flag again, confusing the
    state machine.

If my understanding is correct, fixing the problem should be as easy as
removing the line that sets ESC_DCS from strhandle():

diff --git a/st.c b/st.c
index ef8abd5..b5b805a 100644
--- a/st.c
+++ b/st.c
@@ -1897,7 +1897,6 @@ strhandle(void)
case 'P': /* DCS -- Device Control String */
- term.mode |= ESC_DCS;
case '_': /* APC -- Application Program Command */
case '^': /* PM -- Privacy Message */

I've tried the above patch and it fixes my problem, but I don't know if
it introduces any others.

2 years agoFAQ: fix single-buffer patch
Hiltjo Posthuma [Mon, 1 Jun 2020 12:09:46 +0000 (14:09 +0200)]
FAQ: fix single-buffer patch

rebase against master

2 years agoconfig.def.h: add an option allowwindowops, by default off (secure)
Hiltjo Posthuma [Sat, 30 May 2020 19:56:18 +0000 (21:56 +0200)]
config.def.h: add an option allowwindowops, by default off (secure)

Similar to the xterm AllowWindowOps option, this is an option to allow or
disallow certain (non-interactive) operations that can be insecure or

NOTE: xsettitle() is not guarded by this because st does not support printing
the window title. Else this could be exploitable (arbitrary code execution).
Similar problems have been found in the past in other terminal emulators.

The sequence for base64-encoded clipboard copy is now guarded because it allows
a sequence written to the terminal to manipulate the clipboard of the running
user non-interactively, for example:

printf '\x1b]52;0;ZWNobyBoaQ0=\a'

2 years agoFAQ: add some details about the w3m img hack
Hiltjo Posthuma [Sat, 30 May 2020 19:50:54 +0000 (21:50 +0200)]
FAQ: add some details about the w3m img hack

... and an example patch to switch from double-buffering to a single buffer.

2 years agotiny style fix
Hiltjo Posthuma [Sat, 30 May 2020 19:39:49 +0000 (21:39 +0200)]
tiny style fix

2 years agoPartially add back in "support REP (repeat) escape sequence"
Hiltjo Posthuma [Sat, 30 May 2020 19:34:57 +0000 (21:34 +0200)]
Partially add back in "support REP (repeat) escape sequence"

Add the functionality back in for xterm compatibility, but do not expose the
capability in (yet).

Some notes:

It was reverted because it caused some issues with ncurses in some
configurations, namely when using BSD padding (--enable-bsdpad, BSD_TPUTS) in
ncurses it caused issues with repeating digits.

A fix has been upstreamed in ncurses since snapshot 20200523. The fix is also
backported to OpenBSD -current.

2 years agoCall xsetcursor to set win.cursor in main
Steve Ward [Thu, 21 May 2020 02:24:55 +0000 (22:24 -0400)]
Call xsetcursor to set win.cursor in main

In xsetcursor, remove "DEFAULT(cursor, 1)" because 0 is a valid value.
Increase max allowed value of cursor from 6 to 7 (st extension).

2 years agoRevert "support REP (repeat) escape sequence"
Hiltjo Posthuma [Sat, 16 May 2020 16:06:42 +0000 (18:06 +0200)]
Revert "support REP (repeat) escape sequence"

This reverts commit e8392b282c2eaa28725241a9612804fb55113da4.

There is currently a bug in older ncurses versions (like on OpenBSD) where a
fix for a bug with REP is not backported yet. Most likely in tty/tty_update.c:

Noticed while using lynx (which uses ncurses/curses).
To reproduce using lynx: echo "Z0000000" | lynx -stdin

or using the program:

WINDOW *win;
win = initscr();




return 0;

This prints "ZZZZZZZ" (incorrectly).

2 years agosupport REP (repeat) escape sequence
Avi Halachmi (:avih) [Thu, 14 May 2020 15:18:07 +0000 (18:18 +0300)]
support REP (repeat) escape sequence

The sequence \e[Nb prints the last printed char N (more) times if it's
printable, and it's ignored after newline or other control chars.

This is Ecma-048/ANSI-X3.6 sequence and not DEC VT. It's supported by
xterm, and ncurses uses it when possible, e.g. when TERM is xterm* (and
with this commit also st*).

xterm supports only codepoints<=255, possibly due to internal limits.
We support any value/codepoint which was placed in a cell.

To test:
- tput rep 65 4 -> prints 'AAAA'
- printf "\342\225\246\033[4b" -> prints U+2566 1+4 times.

2 years agoAdd rin terminfo capability
Roberto E. Vargas [Sat, 16 May 2020 10:42:51 +0000 (10:42 +0000)]
Add rin terminfo capability

Tianlin Qu discovered that st is missing rin (scroll back #1 lines).

2 years agoMake shift+wheel behaves as shift+Prev/Next
k0ga [Sat, 16 May 2020 09:48:18 +0000 (09:48 +0000)]
Make shift+wheel behaves as shift+Prev/Next

St uses a very good hack where mouse wheel genereates ^Y and ^E,
that are the same keys that less and vi uses for backward and
fordward scrolling. Scroll, as many terminal emulators, use
shift+Prev/Next for scrolling, but it is also using ^E and ^Y
for scroling, characters that are reserved in the POSIX shell
in emacs mode for end of line and yanking, making scroll unsable
in st.

This patch adds a new hack, making shift+wheel returning the
same sequences than shift+Prev/Next, meaning that scroll or
any other similar program will not be able to differentiate
between them.

2 years agoFix selection: selscroll
Jakub Leszczak [Wed, 6 May 2020 11:36:59 +0000 (13:36 +0200)]
Fix selection: selscroll

2 years agoFix selection: ignore ATTR_WRAP when rectangular selection in getsel
Jakub Leszczak [Wed, 6 May 2020 11:35:53 +0000 (13:35 +0200)]
Fix selection: ignore ATTR_WRAP when rectangular selection in getsel

2 years agoFix selection: selclear in tputc
Jakub Leszczak [Wed, 6 May 2020 11:35:06 +0000 (13:35 +0200)]
Fix selection: selclear in tputc

2 years agocode-style: add fallthrough comment
Hiltjo Posthuma [Sat, 9 May 2020 12:43:31 +0000 (14:43 +0200)]
code-style: add fallthrough comment

Patch by Steve Ward, thanks.

2 years agooptimize column width calculation and utf-8 encode for ASCII
Hiltjo Posthuma [Sat, 9 May 2020 12:03:14 +0000 (14:03 +0200)]
optimize column width calculation and utf-8 encode for ASCII

In particular on OpenBSD and on glibc wcwidth() is quite expensive.
On musl there is little difference.

2 years agofix for incorrect (partial) written sequences when libc wcwidth() == -1
Hiltjo Posthuma [Sat, 9 May 2020 11:56:28 +0000 (13:56 +0200)]
fix for incorrect (partial) written sequences when libc wcwidth() == -1

Fix an issue with incorrect (partial) written sequences when libc wcwidth() ==
-1. The sequence is updated to on wcwidth(u) == -1:

c = "\357\277\275"

but len isn't.

A way to reproduce in practise:

* st -o dump.txt
* In the terminal: printf '\xcd\xb8'
- This is codepoint 888, on OpenBSD it reports wcwidth() == -1.
- Quit the terminal.
- Look in dump.txt (partial written sequence of "UTF_INVALID").

This was introduced in:

" commit 11625c7166b7e4dad414606227acec2de1c36464
Author: <>
Date:   Tue Oct 28 12:55:28 2014 +0100

    Replace character with U+FFFD if wcwidth() is -1

    Helpful when new Unicode codepoints are not recognized by libc."


Remove setting the sequence. If this happens to break something, another
solution could be setting len = 3 for the sequence.

2 years agotiny code-style and typo-fix in comment
Hiltjo Posthuma [Sat, 9 May 2020 11:55:34 +0000 (13:55 +0200)]
tiny code-style and typo-fix in comment

2 years agoauto-sync: draw on idle to avoid flicker/tearing
Avi Halachmi (:avih) [Tue, 26 Feb 2019 20:37:49 +0000 (22:37 +0200)]
auto-sync: draw on idle to avoid flicker/tearing

st could easily tear/flicker with animation or other unattended
output. This commit eliminates most of the tear/flicker.

Before this commit, the display timing had two "modes":

- Interactively, st was waiting fixed `1000/xfps` ms after forwarding
  the kb/mouse event to the application and before drawing.

- Unattended, and specifically with animations, the draw frequency was
  throttled to `actionfps`. Animation at a higher rate would throttle
  and likely tear, and at lower rates it was tearing big frames
  (specifically, when one `read` didn't get a full "frame").

The interactive behavior was decent, but it was impossible to get good
unattended-draw behavior even with carefully chosen configuration.

This commit changes the behavior such that it draws on idle instead of
using fixed latency/frequency. This means that it tries to draw only
when it's very likely that the application has completed its output
(or after some duration without idle), so it mostly succeeds to avoid
tear, flicker, and partial drawing.

The config values minlatency/maxlatency replace xfps/actionfps and
define the range which the algorithm is allowed to wait from the
initial draw-trigger until the actual draw. The range enables the
flexibility to choose when to draw - when least likely to flicker.

It also unifies the interactive and unattended behavior and config
values, which makes the code simpler as well - without sacrificing
latency during interactive use, because typically interactively idle
arrives very quickly, so the wait is typically minlatency.

While it only slighly improves interactive behavior, for animations
and other unattended-drawing it improves greatly, as it effectively
adapts to any [animation] output rate without tearing, throttling,
redundant drawing, or unnecessary delays (sounds impossible, but it

2 years agoreplace exit(3) by _exit(2) in signal handler sigchld()
Jan Klemkow [Wed, 29 Apr 2020 22:10:02 +0000 (00:10 +0200)]
replace exit(3) by _exit(2) in signal handler sigchld()

exit(3) is not async-signal-safe but, _exit(2) is.
This change prevents st to crash and dump core.

2 years agobump version to 0.8.3
Hiltjo Posthuma [Mon, 27 Apr 2020 11:56:25 +0000 (13:56 +0200)]
bump version to 0.8.3

2 years agoUpdate XIM cursor position only if changed
Ivan Tham [Sun, 19 Apr 2020 17:38:39 +0000 (19:38 +0200)]
Update XIM cursor position only if changed

Updating XIM cursor position is expensive, so only update it when cursor
position changed.

2 years agojust remove the EOF message
Hiltjo Posthuma [Sat, 11 Apr 2020 13:45:06 +0000 (15:45 +0200)]
just remove the EOF message

2 years agoAdd st-mono terminfo entry
Roberto E. Vargas Caballero [Sat, 11 Apr 2020 12:46:17 +0000 (14:46 +0200)]
Add st-mono terminfo entry

This entry is intended for monocolor display and it is very
helpful for color haters.

2 years agoconfig.def.h: add a comment for the scroll variable
Hiltjo Posthuma [Sat, 11 Apr 2020 11:56:31 +0000 (13:56 +0200)]
config.def.h: add a comment for the scroll variable

2 years agoFix small typos
Hiltjo Posthuma [Sat, 11 Apr 2020 11:29:48 +0000 (13:29 +0200)]
Fix small typos

2 years agoLaunch scroll program with the default shell
Quentin Rameau [Sat, 11 Apr 2020 10:09:20 +0000 (12:09 +0200)]
Launch scroll program with the default shell

2 years agoUpdate FAQ with the last modifications
Roberto E. Vargas Caballero [Sat, 11 Apr 2020 09:52:58 +0000 (11:52 +0200)]
Update FAQ with the last modifications

2 years agoAdd terminfo entries for backspace mode
Roberto E. Vargas Caballero [Fri, 10 Apr 2020 20:50:23 +0000 (22:50 +0200)]
Add terminfo entries for backspace mode

St used to use backspace as BS until the commit 230d0c8, but due
to general lack of knowledge of lusers, we moved to the most common
configuration in linux to avoid answering the same question 3 times
per month. With the most common configuration we have a backspace
that returns a DEL, and we have a Delete key that doesn't return a
DEL character neither a BS.

When dealing with devices connected using a serial line (or even
with Plan9) it is more common Backspace as BS and Delete as DEL. For
this reason, st is not always the best tool when you talk with a
serial device.

This patch adds new terminfo entries for Backspace as BS and Delete
as DEL. A patch for confg.h is also added, to make easier switch
between both configurations.

2 years agoFix style issue
Roberto E. Vargas Caballero [Fri, 10 Apr 2020 20:26:12 +0000 (22:26 +0200)]
Fix style issue

2 years agottyread: test for EOF while reading tty
Roberto E. Vargas Caballero [Fri, 10 Apr 2020 20:25:46 +0000 (22:25 +0200)]
ttyread: test for EOF while reading tty

When a read operation returns 0 then it means that we arrived to the end of the
file, and new reads will return 0 unless you do some other operation such as
lseek(). This case happens with USB-232 adapters when they are unplugged.

2 years agoAdd support for scroll(1)
Roberto E. Vargas Caballero [Fri, 10 Apr 2020 20:06:32 +0000 (22:06 +0200)]
Add support for scroll(1)

Scroll is a program that stores all the lines of its child and be used in st as
a way of implementing scrollback.

This solution is much better than implementing the scrollback in st itself
because having a different program allows to use it in any other program
without doing modifications to those programs.

2 years agomake argv0 not static, fixes a warning with tcc
Hiltjo Posthuma [Fri, 10 Apr 2020 10:12:43 +0000 (12:12 +0200)]
make argv0 not static, fixes a warning with tcc

Reported by Aajonus, thanks!

2 years agomouseshortcuts: fix custom modifier on release
Avi Halachmi (:avih) [Thu, 2 Apr 2020 08:43:22 +0000 (11:43 +0300)]
mouseshortcuts: fix custom modifier on release

This line didn't work at mshortcuts at config.h:

  /*  mask       button   function  arg       release */
    { ShiftMask, Button2, selpaste, {.i = 0}, 1 },

and now it does work.

The issue was that XButtonEvent.state is "the logical state ... just prior
to the event", which means that on release the state has the Button2Mask
bit set because button2 was down just before it was released.

The issue didn't manifest with the default shift + middle-click on release
(to override mouse mode) because its specified modifier is XK_ANY_MOD, at
which case match(...) ignores any specific bits and simply returns true.

The issue also doesn't manifest on press, because prior to the event
Button<N> was not down and its mask bit is not set.

Fix by filtering out the mask of the button which we're currently matching.

We could have said "well, that's how button events behave, you should
use ShiftMask|Button2Mask for release", but this both not obvious to
figure out, and specifically here always filtering does not prevent
configuring any useful modifiers combination. So it's a win-win.

2 years agoRemove explicit XNFocusWindow
Ivan Tham [Tue, 18 Feb 2020 15:28:47 +0000 (23:28 +0800)]
Remove explicit XNFocusWindow

XCreateIC ICValues default XNFocusWindow to XNClientWindow if not
specified, it can be omitted since it is the same.

From the documentation

> Focus Window
> The XNFocusWindow argument specifies the focus window. The primary
> purpose of the XNFocusWindow is to identify the window that will receive
> the key event when input is composed.
> When this XIC value is left unspecified, the input method will use the
> client window as the default focus window.

2 years agox: fix XIM handling
Quentin Rameau [Sun, 2 Feb 2020 20:47:19 +0000 (21:47 +0100)]
x: fix XIM handling

Do not try to set specific IM method, let the user specify it with

If the requested method is not available or opening fails, fallback to
the default input handler and register a handler on the new IM server
availability signal.

Do the same when the input server is closed and (re)started.

2 years agox: check we still have an XIC context before accessing it
Quentin Rameau [Sun, 2 Feb 2020 16:38:36 +0000 (17:38 +0100)]
x: check we still have an XIC context before accessing it

2 years agox: do not instantiate a new nested list on each cursor move
Quentin Rameau [Sun, 2 Feb 2020 14:38:08 +0000 (15:38 +0100)]
x: do not instantiate a new nested list on each cursor move

2 years agox: move IME variables into XWindow ime embedded struct
Quentin Rameau [Sun, 2 Feb 2020 14:37:29 +0000 (15:37 +0100)]
x: move IME variables into XWindow ime embedded struct

2 years agoIncrease XmbLookupString buffer
Ivan Tham [Sat, 18 Jan 2020 07:52:25 +0000 (15:52 +0800)]
Increase XmbLookupString buffer

Current buffer is too short to input medium to long sentences from IME.
Input with longer text will show the wrong input, taking 64 instead of
32 bytes should be enough for most of the cases. Broken cases before,

Chinese (taken from song 也可以)

Japanese (taken from bootleggers rom quote)

2 years agoupdate FAQ
Hiltjo Posthuma [Sun, 17 Nov 2019 19:04:52 +0000 (20:04 +0100)]
update FAQ

- add common question about the w3m image drawing hack.
- remove some bad advise about $TERM.
- change some links to https.

2 years agoOSC 52 - copy to clipboard: don't limit to 382 bytes
Avi Halachmi (:avih) [Wed, 16 Oct 2019 09:55:53 +0000 (12:55 +0300)]
OSC 52 - copy to clipboard: don't limit to 382 bytes

Strings which an application sends to the terminal in OSC, DCS, etc
are typically small (title, colors, etc) but one exception is OSC 52
which copies text to the clipboard, and is used for instance by tmux.

Previously st cropped these strings at 512 bytes, which for OSC 52
limited the copied text to 382 bytes (remaining buffer space before
base64). This made it less useful than it can be.

Now it's a dynamic growing buffer. It remains allocated after use,
resets to 512 when a new string starts, or leaked on exit.

Resetting/deallocating the buffer right after use (at strhandle) is
possible with some more code, however, it doesn't always end up used,
and to cover those cases too will require even more code, so resetting
only on new string is good enough for now.

2 years agoCSIEscape, STREscape: use size_t for buffer length
Hiltjo Posthuma [Wed, 16 Oct 2019 09:38:43 +0000 (12:38 +0300)]
CSIEscape, STREscape: use size_t for buffer length

2 years agoSTREscape: don't trim prematurely
Avi Halachmi (:avih) [Wed, 16 Oct 2019 09:19:49 +0000 (12:19 +0300)]
STREscape: don't trim prematurely

STRescape holds strings in escape sequences such as OSC and DCS, and
its buffer is 512 bytes.

If the input is too big then trailing chars are ignored, but the test
was off-by-1 such that it took 510 chars instead of 511 (before a
terminating NULL is added).

Now the full size can be utilized.

2 years agobase64dec: don't read out of bounds
Avi Halachmi (:avih) [Wed, 16 Oct 2019 08:17:23 +0000 (11:17 +0300)]
base64dec: don't read out of bounds

Previously, base64dec checked terminating input '\0' every 4 calls to
base64dec_getc, where the latter progressed one or more chars on each
call, and could read past '\0' in the way it was used.

The input to base64dec currently comes only from OSC 52 escape seq
(copy to clipboard), and reading past '\0' or even past the buffer
boundary was easy to trigger.

Also, even if we could trust external input to be valid base64, there
are different base64 standards, and not all of them require padding
to 4 bytes blocks (using trailing '=' chars).

It didn't affect short OSC 52 strings because the buffer is initialized
to 0's, so typically it did stop within the buffer, but if the string
was trimmed to fit (the buffer is 512 bytes) then it did also read past
the end of the buffer, and the decoded suffix ended up arbitrary.

This patch makes base64dec_getc not progress past '\0', and instead
produce fake trailing padding of '='.

Additionally, at base64dec, if padding is detected at the first or
second byte of a quartet, then we identify it as invalid and abort
(a valid quartet has at least two leading non-padding bytes).

2 years agoFix tmux terminfo extensions Se and Ss
Sebastian J. Bronner [Tue, 5 Nov 2019 17:16:39 +0000 (18:16 +0100)]
Fix tmux terminfo extensions Se and Ss

The tmux terminfo extensions Ss and Se are currently specified as
booleans in ``. They should be strings. See
lines 254 and 265.

I have used the values from
for this patch.

2 years agoapply hints before initial mapping (ICCCM)
Ingo Lohmar [Fri, 31 May 2019 20:25:35 +0000 (22:25 +0200)]
apply hints before initial mapping (ICCCM)

For WM_CLASS this is mentioned in the ICCCM docs
(third sentence).

When changing the WM_CLASS from the command line, this is necessary for
window managers to pick it up before applying class-based rules.

2 years agomouse shortcuts: allow using forcemousemod (e.g. shift)
Avi Halachmi (:avih) [Thu, 24 Oct 2019 12:42:07 +0000 (15:42 +0300)]
mouse shortcuts: allow using forcemousemod (e.g. shift)

The recent mouse shurtcuts commits allow customization, but ignore
forcemousemod mask (default: shift) as a modifier, for no good reason
other than following the behavior of the KB shortcuts.

Allow using forcemousemod too, which now can be used to invoke
different shortcuts, though the automatic effect of forcemousemod will
be lost for buttons which use mask with forcemousemod.

E.g. the default is:

static uint forcemousemod = ShiftMask;
{ XK_ANY_MOD,           Button4, ttysend,        {.s = "\031"} },

where ttysend will be invoked for button4 with any mod when not in mouse
mode, and with shift when in mouse mode.

Now it's possible to do this:
{ ShiftMask,            Button4, ttysend,        {.s = "foo"} },
{ XK_ANY_MOD,           Button4, ttysend,        {.s = "\031"} },

Which will invoke ttysend("foo") while shift is held and ttysend("\031")
otherwise. Shift still overrides mouse mode, but will now send "foo".

Previously with this setup the second binding was always invoked
because the forceousemod mask was always removed from the event.

Buttons which don't use forcemousemod behave the same as before.

This is useful e.g. for the scrollback mouse patch, which wants to
configure shift+wheel for scrollback, while keeping the normal behavior
without shift.

2 years agomouse shortcuts: don't hardcode selpaste
Avi Halachmi (:avih) [Thu, 10 Oct 2019 23:26:10 +0000 (02:26 +0300)]
mouse shortcuts: don't hardcode selpaste

Because selpaste is activated on release, a release flag was added to
mouse shortcuts which controls whether activation is on press/release,
and selpaste binding to button2 was moved to config.h .

button1 remains the only hardcoded mouse button - for selection + copy.

2 years agomouse shortcuts: allow override for all shortcuts
Avi Halachmi (:avih) [Thu, 10 Oct 2019 20:42:30 +0000 (23:42 +0300)]
mouse shortcuts: allow override for all shortcuts

Allow forceselmod to override all mouse shortcuts rather than only
selection, and rename it to forcemousemod as it's now more appropriate.

This will affect mouse shortcuts which use mask other than XK_ANY_MOD.

This does not affect the default behavior because the default mouse
shortcuts (wheel) use XK_ANY_MOD, where forceselmod already activated
the override also before this change.

Previously, if a mouse shortcut was configured with a specific mod and
forceselmod was held, then the shortcut did not execute unless the
configured mod included forceselmod.

2 years agomouse shortcuts: allow same functions as kb shortcuts
Avi Halachmi (:avih) [Thu, 10 Oct 2019 20:02:26 +0000 (23:02 +0300)]
mouse shortcuts: allow same functions as kb shortcuts

Previously mouse shortcuts supported only ttywrite.

This required adding an "Arg" function ttysend - which does what the
original mouse shortcuts did.