In the current implementation, when text is written over an image, we
have to cut the entire text line out of the image, regardless of how
long the text is. It doesn't look good, but it was a design choice for
the following reasons:
1) To keep the sixel engine as fast as possible
2) Most applications do not write text on the images anyway
To bring the st terminal in line with other terminals that support
sixels, I have now improved the sixel renderer so that the images can
now have gaps, which allows the text to be printed inside the images.
The changes should not affect performance in normal cases. Only when the
renderer has to deal with the text there might be some performance hits
depending on how many gaps there are in the images.
Patch by Mikhail Kot <to@myrrc.dev>
With some modifications to behave more like xterm (see note below).
Example:
printf '\033[48;2;255:0:0mtest\n'
https://invisible-island.net/xterm/ctlseqs/ctlseqs.html
Some notes:
"CSI Pm m Character Attributes (SGR).
[...]
o xterm allows either colons (standard) or semicolons
(legacy) to separate the subparameters (but after the
first colon, colons must be used).
P2 selects how the terminal draws the background color.
P2 Meaning
0 or 2 (default) Pixel positions specified as 0 are set to the
current background color.
1 Pixel positions specified as 0 remain at their
current color.
Both modes are now supported.
Ref. https://www.vt100.net/docs/vt3xx-gp/chapter14.html
* adds missing function prototype
* move xgetcolor() prototype to win.h (that's where all the other x.c
func prototype seems to be declared at)
* check for snprintf error/truncation
* reduces code duplication for osc 10/11/12
* unify osc_color_response() and osc4_color_response() into a single function
the latter two was suggested by Quentin Rameau in his patch review on
the hackers list.
ref.
https://git.suckless.org/st/commit/8629d9a1da72cc18568a8f146307b0e939b77ebf.html
The general reasoning is that the vim browse patch is very invasive,
has a high level of complexity, and is incompatible with a significant
number of other patches and it complicates further maintenance.
Additionally the patch has its own scrollback mechanism which seemingly
did not work properly - and nobody seems to have complained about this
since the patch was added back in May 2021.
If you want to try out the vim browse patch then I would recommend having
a play around with the patch author's own build that has this patch
integrated:
- https://github.com/juliusHuelsmann/st-history-vim
- https://github.com/juliusHuelsmann/st
Alternatively a tag has been added to this repository that refers to the
last commit that still has the vim browse patch:
- https://github.com/bakkeby/st-flexipatch/tree/VIM_BROWSE_PATCH
* sixel: remove black bars from sixel images
When the images don't fully cover the text cells, black bars are added
to them. This fix removes those bars, but if you need the old behavior,
you can restore it by setting 'sixelremovebars' to zero in config.h
* sixel: fix a potential memory leak
* sixel: improve behavior with text reflow
* sixel: prevent animated gifs from choking the terminal
Animated gifs constantly spawn new images that eventually choke the
terminal because the old animation frames are kept in the image buffer.
This fix removes overlapping images from the image buffer and prevents
them from piling up.
* sixel: add zooming and clipping
* sixel: copying bulk of changes
* sixel: move sixel_parser_parse() and add missing sequences and blocks (#113)
- Move sixel_parser_parse() from tputc() to twrite()
- Add missing 8452, DECSDM, XTSMGRAPHICS and XTWINOPS sequences
- Add more conditional blocks for the scrollback and sync patches
- Remove unused reflow_y from ImageList. It is only used for the
scrollback-reflow patch in st-sx.
* sixel: update vtiden to VT200 family
* sixel: fix scrolling issues inside tmux (#114)
tmux is using the scrolling region and sequence to clear the screen
below the shell prompt. This peculiar behavior caused the tscrollup()
function to be called, which always scrolled the images regardless of
whether they were inside the region or not. So the images moved out of
place whenever the bottom of the screen was cleared. This fix checks
that the images are inside the region before scrolling them.
* sixel: prevent images from being deleted when resizing (#115)
This fixes resizing issues outside of tmux not inside.
* Rewriting tresize logic based on veltza's proposed implementation in PR #115
* tresize: correction for tscrollup call when scrollback patch is used
---------
Co-authored-by: veltza <106755522+veltza@users.noreply.github.com>
The handler for 'S' final character does not check for a private
marker. This can cause a conflict with a sequence called 'XTSMGRAPHICS'
which also has an 'S' final character, but uses the private marker '?'.
Without checking for a private marker, st will perform a scroll up
operation when XTSMGRAPHICS is seen, which can cause unexpected display
artifacts.
ref.
https://git.suckless.org/st/commit/a3f7420310be0fd778ef9fe2abf20edc2d8dc81a.html
It is unclear if it's "required" to do this on RIS, but it's useful when
calling reset(1) after interactive programs have crashed and garbled up
the screen.
FWIW, other terminals do it as well (tested with XTerm, VTE, Kitty,
Alacritty, Linux VT).
ref.
https://git.suckless.org/st/commit/559fdc278681c98470749adb59f01cd071720458.html
Consider the following example:
printf '\e[?7l';\
for i in $(seq $(($(tput cols) - 1))); do printf a; done;\
printf '🙈\n';\
printf '\e[?7h'
Even though MODE_WRAP has been disabled, the emoji appeared on the next
line. This patch keeps wide glyphs on the same line and moves them to
the right-most possible position.
ref.
https://git.suckless.org/st/commit/8abe4bcb41aa7fda0ae00823f6a20271124150db.html
Fixes garbage selections when switching to/from the alternate screen.
How to reproduce:
- Be in primary screen.
- Select something.
- Run this (switches to alternate screen, positions the cursor at the
bottom, triggers selscroll(), and then goes back to primary screen):
tput smcup; tput cup $(tput lines) 0; echo foo; tput rmcup
- Notice how the (visual) selection now covers a different line.
The reason is that selscroll() calls selnormalize() and that cannot find
the original range anymore. It's all empty lines now, so it snaps to
"select the whole line".
ref.
https://git.suckless.org/st/commit/2fc7e532b23e2f820c6b73d352ec7c41fefa45b5.html
Under insert mode, when inserting a normal character in front of
a wide character, the affected region is shifted to the right by
one cell. However, the empty cell is reset as if being a part of a
wide character, causing the following cell being mishandled as a
dummy cell.
To reproduce the bug:
printf '\033[4h' # set MODE_INSERT
printf 妳好
printf '\033[4D'
printf 'x'
printf '\033[4l\n'
Ref.
https://git.suckless.org/st/commit/3a6d6d740110e6ee1b092d05ad746244eedabe4b.html
This patch fixes the following sixel issues:
- The current sixel implementation cleared all cells from the left side
of the image when the image was drawn. The fix only clears the cells
where the image will be drawn.
- The deletion routine didn't work correctly. In certain situations,
it left the image or images undrawn. For example, if the first image
was marked for deletion, it didn't draw the second one.
- The drawing routine caused a high cpu usage, because XCopyArea()
triggered the X server to send the NoExpose event, which caused sixels
to be redrawn and the X server to send another NoExpose event and so
on. This loop caused constant redraw of sixels and high cpu usage.
The fix prevents the X server from sending GraphicsExpose and NoExpose
events.
The patch also adds a control sequence for removing sixels:
Because the sixels are implemented as overlay images, they cannot be
removed by clearing the underlaying cells. Therefore, we need a control
sequence to remove them. I opted to choose ESC[6J as the control
sequence because it is not used and the number refers to sixels. So when
the lf file manager supports sixels [1], you can use the following
minimal scripts to preview images in lf:
previewer:
#!/bin/sh
case "$(readlink -f "$1")" in
*.bmp|*.gif|*.jpg|*.jpeg|*.png|*.webp|*.six|*.svg|*.xpm)
chafa -s "$(($2-3))x$3" -f sixels "$1"
exit 1 ;;
*)
bat "$1" ;;
esac
cleaner:
#!/bin/sh
printf "\033[6J" >/dev/tty
[1] https://github.com/gokcehan/lf/pull/1211
ignore C1 control characters in UTF-8 mode
Ignore processing and printing C1 control characters in UTF-8 mode.
These are in the range: 0x80 - 0x9f.
By default in st the mode is set to UTF-8.
This matches more the behaviour of xterm with the options -u8 or +u8 also.
Also see the xterm resource "allowC1Printable".
Let me know if this breaks something, in most cases I don't think so.
As usual a very good reference is:
https://invisible-island.net/xterm/ctlseqs/ctlseqs.html
Ref.
https://git.suckless.org/st/commit/211964d56ee00a7d46e251cbc150afb79138ae37.html
Add support for DSR response "OK" escape sequence
"VT100 defines an escape sequence [1] called Device Status Report (DSR). When
the DSR sequence received is `csi 5n`, an "OK" response `csi 0n` is returned.
This patch adds that "OK" response.
I encountered this missing sequence when I noticed that fzf [2] would clobber
my prompt whenever completing a find.
To test that ST doesn't currently respond to `csi 5n`, use fzf's shell
extension in ST's repo to complete the path for a file.
my-fancy-prompt $ vim **<tab>
<select a file>
st.c
Select a file with <enter>, and notice that fzf clobbers some or all of your
prompt.
After applying this patch, do the same test as above and notice that fzf has no
longer clobbered your prompt by placing the file name in the correct position
in your command.
my-fancy-prompt $ vim **<tab>
<select a file>
my-fancy prompt $ vim st.c
Thank you for considering my first patch submission.
[1] https://www.xfree86.org/current/ctlseqs.html#VT100%20Mode
[2] https://github.com/junegunn/fzf
"
Patch slightly adapted with input from the mailinglist,
Ref.
https://git.suckless.org/st/commit/f17abd25b376c292f783062ecf821453eaa9cc4c.html
Fixed OSC color reset without parameter->resets all colors
Adapted from (garbled) patch by wim <wim@thinkerwim.org>
Additional notes: it should reset all the colors using xloadcols().
To reproduce: set a different (theme) color using some escape code, then reset
it:
printf '\x1b]104\x07'
Ref.
https://git.suckless.org/st/commit/7e8050cc621f27002eaf1be8114dee2497beff91.html
* fix externalpipein patch
don't close the slave fd, according to the original patch in
https://lists.suckless.org/hackers/2004/17218.html
* externalpipein patch: add example command
press S-C-M to set the terminal background green dynamically.
Replace `printf ...` with `dynamic-colors cycle` command mentioned in
https://lists.suckless.org/hackers/2004/17218.html to cycle though the
available dynamic color themes.
The openurlonclick and scrollback patches are now working together,
so links can be clicked in the scrollback buffer too. This update also
adds url underlining and other improvements to the openurlonclick patch.
The full list of changes in the openurlonclick patch:
- Adds scrollback support
- Adds modkey option
- Better url detection
- Underlines url when the mouse pointer is over a link
- Opens a browser as a background process, so it won't lock the terminal anymore
- Fixes a segmentation fault bug
the array is not accessed outside of base64dec() so it makes sense to
limit it's scope to the related function. the static-storage duration of
the array is kept intact.
this also removes unnecessary explicit zeroing from the start and end of
the array. anything that wasn't explicitly zero-ed will now be
implicitly zero-ed instead.
the validity of the new array can be easily confirmed via running this
trivial loop:
for (int i = 0; i < 255; ++i)
assert(base64_digits[i] == base64_digits_old[i]);
lastly, as pointed out by Roberto, the array needs to have 256 elements
in order to able access it as any unsigned char as an index; the
previous array had 255.
however, this array will only be accessed at indexes which are
isprint() || '=' (see `base64dec_getc()`), so reducing the size of the
array to the highest printable ascii char (127 AFAIK) + 1 might also be
a valid strategy.
ref. https://git.suckless.org/st/commit/ef0551932fb162f907b40185d2f48c3b497708ee.html
Overtyping the first half of a wide character with the
second half of a wide character results in display garbage.
This is because the trailing dummy is not cleaned up.
i.e. ATTR_WIDE, ATTR_WDUMMY, ATTR_WDUMMY
Here is a short script for demonstrating the behavior:
#!/bin/sh
alias printf=/usr/bin/printf
printf こんにちは！; sleep 2
printf '\x1b[5D'; sleep 2
printf へ; sleep 2
printf ' '; sleep 2
echo
Ref.
- https://git.suckless.org/st/commit/65f1dc428315ae9d7f362e10c668557c1379e7af.html
In the current implementation, the slave PTY (assigned to the variable
`s') is always closed after duplicating it to file descriptors of
standard streams (0, 1, and 2). However, when the allocated slave PTY
`s' is already one of 0, 1, or 2, this causes unexpected closing of a
standard stream. The same problem occurs when the file descriptor of
the master PTY (the variable `m') is one of 0, 1, or 2.
In this patch, the original master PTY (m) is closed before it would
be overwritten by duplicated slave PTYs. The original slave PTY (s)
is closed only when it is not one of the standarad streams.
Ref. https://git.suckless.org/st/commit/1d3142da968da7f6f61f1c1708f39ca233eda150.html
#if SIXEL_PATCH
case 't':
/* TODO should probably not be hard-coded */
ttywrite(";420;720t", 10, 1);
break;
#endif // SIXEL_PATCH
This would result in printing ";420;720t" when exiting neovim.
Without this code a line is written to standard err instead:
erresc: unknown csi ESC[23;0t
The ttywrite was added as part of this commit:
- b50be8225d
which states:
> When a S or T CSI escape was encountered, the lines which were scrolled
> away would be deleted from the scrollback buffer. This has been
> corrected - the lines are now preseved.
>
> This fixes a bug where issuing `clear` followed by `lsix` would cause
> the line on which the `lsix` was issued to disappear from the scrollback
> buffer.
>
> Note that the line may scroll out of view and thus dissapear, but it
> will now be preserved in the scrollback buffer.
Given that we could not reproduce the above bug without the ttywrite in
this case I am not convinced that this is actually needed. Leaving this
here in case this comes up again in the future.
