This patch enables bracketed paste mode (
http://invisible-island.net/xterm/ctlseqs/ctlseqs.html#Bracketed%20Paste%20Mode
).
It's mainly useful for text editors to disable line wrapping and auto
indentation when text is being pasted, rather than typed from keyboard.
On the emulator side, it is supported by at least xterm, urxvt,
gnome-terminal, putty, iterm2; and I have a patch for konsole.
On the application side, vim can be configured easily to handle this, and
I have pending patches for mcedit and joe. Probably many others also
support it.
* Button number in X10 mode:
I believe the button - 1 came from "C b is button - 1" from [0].
However, above this section, it states
"Normally, parameters (such as pointer poisition and button number)
for all mouse tracking escape sequences generated by xterm encode
numeric parameters in a single character as value+32. For example, !
specifies the value 1."
Also, from the description of SGR,
"The encoded button value in this case does not add 32 since that
was useful only in the X10 scheme for ensuring that the byte
containing the button value is a printable code."
This suggests that we should still add 32 to the button value when in
MODE_MOUSEX10.
* No button release reporting in X10 mode:
"X10 compatibility mode sends an escape sequence only on button press,
encoding the location and the mouse button pressed."
* Fix MODE_MOUSEMOTION:
Currently, motion reporting is skipped when oldbutton == 3
(corresponding to no button being pressed). However, oldbutton is
only set on a button press, which will never be 3.
[0]: http://invisible-island.net/xterm/ctlseqs/ctlseqs.html
\a is the character for bell, and st is only marking the window as urgent
if it is not active. This patch adds an audible bell which can be disable
with bellvolume variable.
Signed-off-by: Christoph Lohmann <20h@r-36.net>
The alternate screen is not properly initialized when st starts. To see
this, set defaultbg in config.h to anything other than 0 (for example, swap
defaultfg and defaultbg), and run:
./st -e sh -c 'tput smcup; read'
You should see that the top-left 80x24 rectangle is black (or whatever
colorname[0] is), while the rest of the screen (if any) has the desired
colorname[defaultbg] color.
The attached patch fixes this by initializing term.c.attr in tnew() before
calling tresize(). It also removes the unnecessary xcalloc() calls, which
misled me on this bug hunt since it is really tclearregion() which
initializes term.lines and term.alt in tresize().
Hello.
I reviewed and tested commit 7e3cff3, and made a patch that fixes some
problems in it.
1. There's a semicolon after an if statement, which is obviously a
typo.
2. The current way of calculating text position in "xdraws" yields
inconsistent results in some cases. This is due to the use of
"font->width", which varies. Instead, "xw.cw" has to be used as the
character width.
Sincerely,
Eon
tdefcolor() returns -1 on error, while its return type is
unsigned long. At the same time, line 1724 and 1731 are checking the
positivity of its unsigned return value.
Colors definition can be changed using a OSC sequence, so
we have to reload them if we want be sure all the colors
are the correct.
Could be desirable free the colors allocated due to rgb
colors and inverse colors (XftColorAllocValues in xdraws),
but it is impossible due we use the same structure for all
of them.
This patch uses the bit 24 in the color descriptor as an indicator
of RGB color, so we can take the values and generating the XftColour
directly in xdraws.
I made a patch that improves the performance of font caching mechanism.
This is based on a funny behaviour of FontConfig: it was handling
FcCharSet in a somewhat unexpected way.
So, we are currently adding "a character" to a new FcCharSet, and then
add it to a FcPattern. However, if we toss the FcPattern to FontConfig,
it loads the entire language(charset) that contains the character we
gave. That is, we don't always have to load a new font for each unknown
character. Instead, we can reused cached fonts, and this significantly
reduces the number of calls to extremely slow FontConfig matching
functions.
One more thing. I found that, in libXft, there's a function called
XftCharExists. XftCharIndex internally calls this function, and
does more stuffs if the character does exist. Since the returned index
is never used in st, we should call XftCharExists instead of
XftCharIndex. Please note that I already made this change in the patch.
To be more specific, now tty creation is delayed until X window is
actually mapped; last ConfigureNotify before mapping determines
initial tty size.
Please report problems if there are any.
There were two problems with match denfinition.
1) There was a forward declaration in the form:
static inline bool match(uint, uint);
but later the function was defined as:
inline bool
match(uint mask, uint state) {
This causes that there were two different functions in the code, one local
and inline, and other inline but extern. All was working without problems
due to we were using -Os, and the compiler was using the extern definition
and it was no expanding the static declaration. If you removed the -Os flag,
then you got linker errors due it was no able to find the static definition
of the static declaration.
2) The mask checking was incorrect because we were doing the test:
(state & mask) != state
and this test only was saying that at least all the enabled bits of state
were enabled also in mask, but no all the possible bits in mask. This was
the origin of the bug reported by Xavier Cartron, where he said it was
possible activated some shortcuts with some of the modifiers defined in the
config.h file.
draw is the function which update the Xwindow with the information st has,
and it is designed in a way that it must be called once in the main loop
(run function), and calling it in other places it is a waste of time.
The way st knows if there is a selection activated is checking if sel.ob.x
is equal to -1. In some parts of the code the way of disabling the selection
was only setting it to -1, but after it you can't be sure if the selection
is clearing from the terminal representation, because it is necessary mark
all the lines affected by the selection as dirty. Already there is a functon
which perform this task, selclear.
We're now clearing empty areas with spaces, so there is no point to check
if character contains non-empty string.
Signed-off-by: Christoph Lohmann <20h@r-36.net>
The commit b78c5085f7 changed the st behaviour enabling BCE capability,
that means erase regions using background color. Problem comes when you
clear a region with a selection, because in this case the real mode of the
Glyph is not the value of term.line[y][x], due in drawregion we had enabled
the ATTR_REVERSE bit.
Signed-off-by: Christoph Lohmann <20h@r-36.net>
Now double-click+dragging automatically snaps both ends to word boundaries
(unless on series of spaces), and triple-click selects whole lines.
As a side effect, snapping now occurs on button press, not button release
like it previously was, but I hope that won't be inconvenient for anyone.
Signed-off-by: Christoph Lohmann <20h@r-36.net>
Now, when you are selecting a region, you will get all empty lines that happen
to be in it, including trailing ones. Last line terminator is omitted as it previously
was, though.
Signed-off-by: Christoph Lohmann <20h@r-36.net>
To have a more visible cursor on unfocused windows this patch makes st draw a
rectangle around the terminal cell.
Thanks Mark Hills <mark@xwax.org> for the suggestion!
The copying and pasting in the terminald and GUI world is flawed. Due to the
discussion on the mailinglist it seems that sending '\n' is what GUIs expect
and '\r' what terminal applications want. St now implements that behaviour.
People sending me patches against strange revisions and basing on their own
revisions make me having to reapply them. Then such errors appear.
Thanks Alexander Sedov <alex0player@gmail.com> for noticing this.
The specified font[] pattern need not have a medium weight. It could be
specified as a number too or have a different weight other than medium.
Signed-off-by: Christoph Lohmann <20h@r-36.net>
Keypad will generate keycodes when keypad application mode is enabled. It
can cause problems with some programs like vi, which operates in such
mode.
This patch change by default don't generate the keycodes never, but this
behaviour can be changed using the combination Alt + NumLock.
---
config.def.h | 34 ++++++++++++++++++----------------
st.c | 17 +++++++++++++++--
2 files changed, 33 insertions(+), 18 deletions(-)
This patch apply the same code for shortcuts that it is used now for defined
keys. So it is possible use now XK_NO_MOD and XK_ANY_MOD for defining shortcuts.
---
st.c | 35 ++++++++++++++++++++---------------
1 file changed, 20 insertions(+), 15 deletions(-)
XK_NO_MOD match a key without modifiers and XK_ANY_MOD match a key does not
matter what modifiers are pressed to. Like they are mask the best value for
XK_ANY_MOD is all the bits to 1, so the and with any state will be equal to
the state. This also imply that is necessary check the case for XK_NO_MOD
(no modifiers at all) with some modifier in state, and the inverse
(some mask different to XK_ANY_MOD or XK_NO_MOD and no modifiers in state).
---
st.c | 12 +++++++-----
1 file changed, 7 insertions(+), 5 deletions(-)
When Shift + Tab is pressed X server send the event XK_ISO_Left_Tab with
ShiftMask, so this is the entry we need in config.def.h
This patch also revert the previous patch for this issue because it breaks
the keyboard.
---
config.def.h | 2 +-
st.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
Some keys were in the Key array while others were hardcoded in
kpress().This cause some problems with some keys which can generate more of
one string based in the configuration of the terminal.
---
config.def.h | 70 ++++++++++++++++++++++++++++++++++++++++-----------------
st.c | 71 +++++++++++++++++++++++++---------------------------------
2 files changed, 79 insertions(+), 62 deletions(-)
Shift + Insert is used like a hot key for paste the selection, so it is more
logical move it to shortcut array instead of having special code for it.
---
config.def.h | 1 +
st.c | 13 +++----------
2 files changed, 4 insertions(+), 10 deletions(-)
Keypad mode is used for detecting when keys in the auxiliary keypad are
pressed, while cursor mode is used for detecting when a cursor is pressed,
but they are different modes.
St was mixing both modes and DECPAM and DECPNM modified the cursor mode, and
this was incorrect.
---
st.c | 5 +++--
st.info | 4 ++--
2 files changed, 5 insertions(+), 4 deletions(-)
These sequences will be never implemented and in this moment they are
generating a lot of noise.
---
st.c | 14 +++++++-------
1 file changed, 7 insertions(+), 7 deletions(-)
DECARM modify the auto repeat settings in the keyboard, and since we can not
modify this setting in the Xserver the best solution is only ignore it.
---
st.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
VPR stands for Move cursor down a number of rows, and the code was moving
the cursor up instead of moving it down.
---
st.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
DECOM sequence allows to the user defines a new home position. The home
position is used as base for all the movement commands except HVP and
VPA. It is important notice than DECSLM moves cursor to absolute position
0,0.
---
st.c | 33 ++++++++++++++++++++++++++-------
1 file changed, 26 insertions(+), 7 deletions(-)
After terminal reset saved terminal position is reset to 0, allowing know
where cursor will go in next restore cursor operation.
---
st.c | 2 ++
1 file changed, 2 insertions(+)
Since relational expresions are always evaluated to 0 or 1, we can use
bitwise xor operator instead of using more complex boolean expressions.
---
st.c | 14 +++++++-------
1 file changed, 7 insertions(+), 7 deletions(-)
Sequences like DECSC, DECRC, ESC [?1047l or ESC [?1047h save and restore
cursor attributes, than taken from vt100 manual are:
Save Cursor (DECSC) ESC 7
===========================
Saves the following in terminal memory.
- cursor position
- graphic rendition
- character set shift state
- state of wrap flag
- state of origin mode
Restore Cursor (DECRC) ESC 8
===========================
Restores the states described for (DECSC) above. If none of these
characteristics were saved, the cursor moves to home position; origin
mode is reset; no character attributes are assigned; and the default
character set mapping is established.
This implies that hide attribute of the cursor should not be saved/restored
in these sequences. The best way to fix this problem is moving hide
attribute into the terminal mode, instead of having it in the cursor state.
---
st.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
write can write less bytes than we request, so it is necessary check the
return value, in case of error print a message and don't continnue writing
in the file.
---
st.c | 39 ++++++++++++++++++++++++++++++---------
1 file changed, 30 insertions(+), 9 deletions(-)
This sequence was used by DEC personal in to for verifying the screen adjust
of terminals. It is the unique test sequence implemented by all the
emulators, and I think it is because they want be conforms with vttest which
uses this sequence in some tests.
---
st.c | 31 +++++++++++++++++++++++--------
1 file changed, 23 insertions(+), 8 deletions(-)
If vt100_0 is a automatic variable then it is initializated in each call to
tsetchar, but if the variable is static it is initializated only in compile
time.
---
st.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
Do not send NUL character in the identification (use (sizeof(VT102ID) - 1),
and finish the sequence once you execute it.
---
st.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
These sequences are used by the host in order to can detect which kind of
terminal is connected. St will answer like a vt102 terminal with this patch.
---
st.c | 9 +++++++++
1 file changed, 9 insertions(+)
Non handled codes must be ignored, except in graphic mode. Also STR
sequences have higher priority than control codes, so they must be handled
before of them.
---
st.c | 160 ++++++++++++++++++++++++++++++++++++------------------------------
1 file changed, 87 insertions(+), 73 deletions(-)
SI and SO allows change the G0 and G1 selection. This implementation is not
full vt100 compatible, but it is complatible with linux virtual terminal
implementation. For full vt100 compatibility we need remake a lot of stuff
relate to the different charmaps.
---
st.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
Add the documentation from the vt100 manual programmer:
Control Octal Action Taken
Character Code
-------------------------------------------
NUL 000 Ignored on input (not stored in input buffer;
see full duplex protocol).
ENQ 005 Transmit answerback message.
BEL 007 Sound bell tone from keyboard.
BS 010 Move the cursor to the left one character position,
unless it is at the left margin,
in which case no action occurs.
HT 011 Move the cursor to the next tab stop,
or to the right margin if no further tab stops
are present on the line.
LF 012 This code causes a line feed or
a new line operation. (See new line mode).
VT 013 Interpreted as LF.
FF 014 Interpreted as LF.
CR 015 Move cursor to the left margin on the current line.
SO 016 Invoke G1 character set, as designated by SCS
control sequence.
SI 017 Select G0 character set, as selected by ESC ( sequence.
XON 021 Causes terminal to resume transmission.
XOFF 023 Causes terminal to stop transmitted all codes
except XOFF and XON.
CAN 030 If sent during a control sequence, the sequence is
immediately terminated and not executed. It also causes
the error character to be displayed.
SUB 032 Interpreted as CAN.
ESC 033 Invokes a control sequence.
DEL 177 Ignored on input (not stored in input buffer).
--------------------------------------------
---
st.c | 28 +++++++++++++++++++---------
1 file changed, 19 insertions(+), 9 deletions(-)
sequences, so we have to support escape sequences in escape sequences that
escape sequences in escape sequences – setting a title won't notify you
anymore.
Taken from vt100 manual programmer:
Control characters (codes \0 to \37 inclusive) are specifically
excluded from the control sequence syntax, but may be embedded
within a control sequence. Embedded control characters are executed
as soon as they are encountered by the VT100. The processing of the
control sequence then continues with the next character received.
---
st.c | 68 +++++++++++++++++++++++++++++++++---------------------------------
1 file changed, 34 insertions(+), 34 deletions(-)
Taken from vt100 programmer manual:
Control characters have values of \000 - \037, and \177. The control
characters recognized by the VT100 are shown in Table 3-10. All
other control codes cause no action to be taken.
We have to take attention when we are using alternate charset, because in
this cases they are not used as control characters.
---
st.c | 20 ++++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)
If -f options is enabled then tputc() writes all the data to a file. Actual
code assumes that all the strings in 'c' parameters have always 1 byte
length, but this is not always true, because due to utf-8 encoding some
characters can have a diferent length. So it is necessary pass string length
to tputc in order it can call to write() correctly.
---
st.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
This sequence lock/unlock the keyboard ignoring all the key pressing events
from X server.
---
st.c | 8 +++++++-
1 file changed, 7 insertions(+), 1 deletion(-)
tsetreset() is called when it is necessary a full initialization of the
terminal, so it also should clean the full X window and not only the
terminal content. It is necessary change the order of the
initialization in main(), and put xinit before of tnew(), because tnew()
calls to tsetreset(), and this can cause a call to xreset() with
incorrect values.
---
st.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
Some times the size after a resizing is not an exact multiply of a number of
characters, so redrawn the screen using the lines and columns of the neww
size can cause that some old graphics keep in the screen. Solution is clean
all the windows with the background color.
---
st.c | 3 +++
1 file changed, 3 insertions(+)
st selection don't insert in the selection position whose value is not
set. This is correct for the positions in the end of the line, but cause
some problems in the beginning. For example echo -e 'a\tb' will print in the
screen:
a b
but after selecting and copying in some place you get:
ab
because positions from 1 to 7 don't have any value. This patch deals all
positions without value as blank (even at the end of the line).
---
st.c | 17 ++++++++++-------
1 file changed, 10 insertions(+), 7 deletions(-)
malloc and realloc are called through xmalloc and xrealloc, so calloc should
be called through xcalloc.
---
st.c | 13 +++++++++++--
1 file changed, 11 insertions(+), 2 deletions(-)
In previous commits draw was removed from all the X events, but I forgot do
it in resize.
---
st.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
After the commit named "Remove timeout in the main loop", selection is not
working in the proper way. After selecting something, press mouse button in
a line outside of selection causes an incorrect highlight. This patch fix
the problem forcing a draw after the press event, but this is only a fast
hack. Real solution means rewriting selection code.
---
st.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
draw() runs over all lines of the screen and renders only the dirty lines,
this avoids render lines which are not modified since last draw() call. In
this moment the main loop is something like:
- Wait something to read from file descriptors
- Read from pseudo tty
- Call draw() for rending
- Read X events
This cause the problem that all the X events that have to update the screen
have to call draw() (because draw() is called before of X events handling),
so you can have multiples renderings in only one iteration, that will waste
a lot of resources.
This patch change the main loop to:
- Wait something to read from file descriptors
- Read from pseudo tty
- Read X events
- Call draw() for rending
So X events don't have to worry about rendering, because draw() is called
after them.
The only place where draw is called outside of the main loop is in redraw(),
but it is necessary for getting a good tput flash.
---
st.c | 29 ++++++-----------------------
1 file changed, 6 insertions(+), 23 deletions(-)
The main loop waits until there is some data to read in file descriptors of
the X server or the pseudo tty. But it uses a timeout in select(), which
causes that st awake each 20 ms, even it doesn't have something to do. This
patch removes this problem removing the timeout, which is not needed.
---
TODO | 1 -
st.c | 27 +++------------------------
2 files changed, 3 insertions(+), 25 deletions(-)
It is necessary call to XSync if you want a good tput flash, because in
other way you can not be sure that white screen will be shown.
---
st.c | 1 +
1 file changed, 1 insertion(+)
XdbeQueryExtension() tells to the caller if the Xdbe extension is present in
the X server, so it should be called for sanity. But like is said in
XdbeQueryExtension(3):
No other Xdbe functions may be called before this function. If a
client violates this rule, the effects of all subsequent Xdbe calls
that it makes are undefined.
it is mandatory call this function.
---
st.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
By default text files are line buffered, and this means that -f option will
not write the line until a \n is printed. This is not very useful for
debugging, so a call to fflush was added. This patch substitute this call
(which will be done by each character painted) by the full remove of the
buffering in the file.
---
st.c | 12 +++++-------
1 file changed, 5 insertions(+), 7 deletions(-)
If malloc or realloc fail they return NULL. Theorically this condition
should be tested in the code, but it's a strange condition today (basically
if this is hapenning thenyou have a big problem), and even Linux never returns
NULL in the default configuration (only if the process don't have room in
the space address, something a bit impossible in the case of st). But stis
enough small for being executed in low resources computers where this can be
a real problem. So the easy way is creating a wrappers function for them and
call to die in case of error.
---
st.c | 44 +++++++++++++++++++++++++++++++-------------
1 file changed, 31 insertions(+), 13 deletions(-)
st marks the active selection using reverse colors in the box selection, but
once that another window becomes owner of the selection, it is very
confusing that st keeps highlight the old selection. Usually terminal
emulators remove the highlight when it is not valid anymore.
X sends a SelectionClear event in this situation, so we only have to add a
callback which unhighlight the selectin box.
---
st.c | 9 +++++++++
1 file changed, 9 insertions(+)
Some programs use the alternative screen (vi, less, ...), whose
content is different of the main screen. If you select text in one of
the screen, you don't wait the box selection is painted in the other
screen, so it is necessary check if the selection was done in the same
screen we are going to paint. Before to this commit, you could do
something like:
$ LESS="" ls | less
(select some code)
q
and selection box remains drawing in the main screen, but the content
of selection keeps text of the alternate screen.
---
st.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
When it is called DECSCNM all lines become dirty, because it is necessary
redraw all lines for getting the new colors. It is easy see the problem
running 'echo ^[[?5h'.
In order to get a correct flash when running tput flash is necessary wait
after DECSCNM, until the changes are displayed, because in other case the
switch between reverse on/reverse off will be too much fast and nothing will
happen.
---
st.c | 12 +++++++++++-
1 file changed, 11 insertions(+), 1 deletion(-)
This is a theorical feature listed in http://st.suckless.org/goals. All the
input/output of the terminal will be written to a file, which can be very
useful for debugging, and also allow interconnect st to other process
through named pipes.
---
st.1 | 6 ++++++
st.c | 14 +++++++++++++-
2 files changed, 19 insertions(+), 1 deletion(-)
SM and RM can receive multiple parameters, but the code only was accepting
only one. This patch join the code of set and reset modes (SM and RM) in a
common function and uses a loop which deals with all the arguments of the
sequence. This patch improves xterm and vt100 compability.
---
st.c | 180 ++++++++++++++++++++++++++++--------------------------------------
1 file changed, 76 insertions(+), 104 deletions(-)
This sequence performs "Cursor Forward Tabulation <n> tab stops", which
although is not present in vt100 or vt102, xterm accepts it.
---
st.c | 5 +++++
1 file changed, 5 insertions(+)
This sequence clears tab stops in the terminal. If the argument is not present
or is zero, then removes the tab stop of the current horizontal position. If
the argument is 3 then removes all the tab stops of the terminal. It was
necessary modify the terminfo entry tbc, because it has \E[2g instead of the
correct \E[3g.
Signed-off-by: Roberto E. Vargas Caballero <k0ga@shike2.com>
---
st.c | 12 ++++++++++++
st.info | 2 +-
2 files changed, 13 insertions(+), 1 deletion(-)
This sequence adds a new tab stop in the current horizontal position. This
means that tputtab must be look for the next tab stop in the tabs array
instead of using a hard coded value offset. Also, CHT sequence XXX message
is removed because it is not a vt10x sequence (as far as I know it is a
vt50x sequence), and it is not implemented by linux virtual terminal neither
by xterm.
Signed-off-by: Roberto E. Vargas Caballero <k0ga@shike2.com>
---
st.c | 12 +++++++++---
1 file changed, 9 insertions(+), 3 deletions(-)
Tabs stop are simulated in st using a fixed size of 8, always, without be
worried about sequences changing the tab stops. A user can put a tab stop in
each horizontal position of the screen, so we need at least one flag for
each column of the screen. In the same way as dirty flags is used for the
rows, it is used a bool dinamic array.
Signed-off-by: Roberto E. Vargas Caballero <k0ga@shike2.com>
---
st.c | 22 +++++++++++++++++++---
1 file changed, 19 insertions(+), 3 deletions(-)
* add a timeout value (SELECT_TIMEOUT) of 20ms in the select() call
* wait at least 20ms (DRAW_TIMEOUT) between draw() calls
* only copy dirty lines from the buffer to the screen
what draw() does:
* clears dirty lines in the buffer
* draws the longest same-attributes string of each
dirty line to the buffer with multiple xdraws() call
* copies the current dirty line from buffer to the screen with a single
xcopy() call
this changeset makes st run ~10x faster.
results in a lot of simplification :
- no more dirty flags (perf are good enough).
- no more ugly gfx call in emulation functions.
LINESPACE removed from config.h.
BORDER is now handled correctly.
cursor keys are handled in kpress according to the Application Mode (DECPAM).
define & enum were renamed.
tcursor() is now tmovecursor() which is more correct.
tcpos() is now tcursor(), as DECSC is also supposed to save attributes.
capnames are indicated whenever possible.
Currently:
alsamixer looks fine, totally usable.
ncmpc is almost ok.
emacs looks like shit.
