path: root/runtime/doc/terminal.txt

*terminal.txt*	For Vim version 8.1.  Last change: 2018 May 17


		  VIM REFERENCE MANUAL	  by Bram Moolenaar


Terminal window support				*terminal* *terminal-window*


The terminal feature is optional, use this to check if your Vim has it: >
	echo has('terminal')
If the result is "1" you have it.


1. Basic use			|terminal-use|
      Typing				|terminal-typing|
      Size and color			|terminal-size-color|
      Syntax				|:terminal|
      Resizing				|terminal-resizing|
      Terminal Modes			|Terminal-mode|
      Cursor style			|terminal-cursor-style|
      Session				|terminal-session|
      Special keys			|terminal-special-keys|
      Unix				|terminal-unix|
      MS-Windows			|terminal-ms-windows|
2. Terminal communication	|terminal-communication|
      Vim to job: term_sendkeys()	|terminal-to-job|
      Job to Vim: JSON API		|terminal-api|
      Using the client-server feature	|terminal-client-server|
3. Remote testing		|terminal-testing|
4. Diffing screen dumps		|terminal-diff|
      Writing a screen dump test for Vim  |terminal-dumptest|
      Creating a screen dump		  |terminal-screendump|
      Comparing screen dumps		  |terminal-diffscreendump|
5. Debugging			|terminal-debug|
      Starting				|termdebug-starting|
      Example session			|termdebug-example|
      Stepping through code		|termdebug-stepping|
      Inspecting variables		|termdebug-variables|
      Other commands			|termdebug-commands|
      Prompt mode			|termdebug-prompt|
      Communication			|termdebug-communication|
      Customizing			|termdebug-customizing|

{Vi does not have any of these commands}
{only available when compiled with the |+terminal| feature}
The terminal feature requires the |+multi_byte|, |+job| and |+channel| features.

==============================================================================
1. Basic use						*terminal-use*

This feature is for running a terminal emulator in a Vim window.  A job can be
started connected to the terminal emulator. For example, to run a shell: >
     :term bash

Or to run build command: >
     :term make myprogram

The job runs asynchronously from Vim, the window will be updated to show
output from the job, also while editing in another window.


Typing ~
							*terminal-typing*
When the keyboard focus is in the terminal window, typed keys will be sent to
the job.  This uses a pty when possible.  You can click outside of the
terminal window to move keyboard focus elsewhere.

CTRL-W can be used to navigate between windows and other CTRL-W commands, e.g.:
	CTRL-W CTRL-W	move focus to the next window
	CTRL-W :	enter an Ex command
See |CTRL-W| for more commands.

Special in the terminal window:			*CTRL-W_.*  *CTRL-W_N* 
	CTRL-W .	send a CTRL-W to the job in the terminal
	CTRL-W CTRL-\	send a CTRL-\ to the job in the terminal
	CTRL-W N	go to Terminal-Normal mode, see |Terminal-mode|
	CTRL-\ CTRL-N   go to Terminal-Normal mode, see |Terminal-mode|
	CTRL-W " {reg}  paste register {reg}		*CTRL-W_quote*
			Also works with the = register to insert the result of
			evaluating an expression.
	CTRL-W CTRL-C	ends the job, see below |t_CTRL-W_CTRL-C|

See option 'termwinkey' for specifying another key instead of CTRL-W that
will work like CTRL-W.  However, typing 'termwinkey' twice sends 'termwinkey'
to the job.  For example:
	'termwinkey' CTRL-W    move focus to the next window
	'termwinkey' :	       enter an Ex command
	'termwinkey' 'termwinkey' send 'termwinkey' to the job in the terminal
	'termwinkey' .	       send 'termwinkey' to the job in the terminal
	'termwinkey' CTRL-\    send a CTRL-\ to the job in the terminal
	'termwinkey' N	       go to terminal Normal mode, see below
	'termwinkey' CTRL-N    same as CTRL-W N
	'termwinkey' CTRL-C    same as |t_CTRL-W_CTRL-C|
							*t_CTRL-\_CTRL-N*
The special key combination CTRL-\ CTRL-N can be used to switch to Normal
mode, just like this works in any other mode.
							*t_CTRL-W_CTRL-C*
CTRL-W CTRL-C can be typed to forcefully end the job.  On MS-Windows a
CTRL-BREAK will also kill the job.

If you type CTRL-C the effect depends on what the pty has been configured to
do.  For simple commands this causes a SIGINT to be sent to the job, which
would end it.  Other commands may ignore the SIGINT or handle the CTRL-C
themselves (like Vim does).

To change the keys you type use terminal mode mappings, see |:tmap|.
These are defined like any mapping, but apply only when typing keys that are
sent to the job running in the terminal.  For example, to make F1 switch
to Terminal-Normal mode: >
   tnoremap <F1> <C-W>N
You can use Esc, but you need to make sure it won't cause other keys to
break (cursor keys start with an Esc, so they may break): >
   tnoremap <Esc> <C-W>N
   set notimeout ttimeout timeoutlen=100

You can also create menus similar to terminal mode mappings, but you have to
use |:tlmenu| instead of |:tmenu|.

<							*options-in-terminal*
After opening the terminal window and setting 'buftype' to "terminal" the
TerminalOpen autocommand event is triggered.  This makes it possible to set
options specifically for the window and buffer.  Example: >
   au TerminalOpen * if &buftype == 'terminal' | setlocal bufhidden=hide | endif
The <abuf> is set to the terminal buffer, but if there is no window (hidden
terminal) then setting options will happen in the wrong buffer, therefore the
check for &buftype in the example.

Mouse events (click and drag) are passed to the terminal.  Mouse move events
are only passed when Vim itself is receiving them.  For a terminal that is
when 'balloonevalterm' is enabled.


Size and color ~
							*terminal-size-color*
See option 'termwinsize' for controlling the size of the terminal window.
(TODO: scrolling when the terminal is larger than the window)

The job running in the terminal can change the colors.  The default foreground
and background colors are taken from Vim, the Normal highlight group.

For a color terminal the 'background' option is used to decide whether the
terminal window will start with a white or black background.

To use a different color the Terminal highlight group can be used, for
example: >
    hi Terminal ctermbg=lightgrey ctermfg=blue guibg=lightgrey guifg=blue
<
							*g:terminal_ansi_colors*
In GUI mode or with 'termguicolors', the 16 ANSI colors used by default in new
terminal windows may be configured using the variable
`g:terminal_ansi_colors`, which should be a list of 16 color names or
hexadecimal color codes, similar to those accepted by |highlight-guifg|.  When
not using GUI colors, the terminal window always uses the 16 ANSI colors of
the underlying terminal.
The |term_setansicolors()| function can be used to change the colors, and
|term_getansicolors()| to get the currently used colors.


Syntax ~

:[range]ter[minal] [options] [command]			*:ter* *:terminal*
			Open a new terminal window.

			If [command] is provided run it as a job and connect
			the input and output to the terminal.
			If [command] is not given the 'shell' option is used.
			if [command] is NONE no job is started, the pty of the
			terminal can be used by a command like gdb.

			If [command] is missing the default behavior is to
			close the terminal when the shell exits.  This can be
			changed with the ++noclose argument.
			If [command] is present the default behavior is to
			keep the terminal open in Terminal-Normal mode.  This
			can be changed with the ++close argument.

			A new buffer will be created, using [command] or
			'shell' as the name, prefixed with a "!".  If a buffer
			by this name already exists a number is added in
			parentheses.  E.g. if "gdb" exists the second terminal
			buffer will use "!gdb (1)".

			If [range] is given the specified lines are used as
			input for the job.  It will not be possible to type
			keys in the terminal window.  For MS-Windows see the
			++eof argument below.

						*term++close* *term++open*
			Supported [options] are:
			++close		The terminal window will close
					automatically when the job terminates.
			++noclose	The terminal window will NOT close
					automatically when the job terminates.
			++open		When the job terminates and no window
					shows it, a window will be opened.
					Note that this can be interruptive.
				The last of ++close, ++noclose and ++open
				matters and rules out earlier arguments.

			++curwin	Open the terminal in the current
					window, do not split the current
					window.  Fails if the current buffer
					cannot be |abandon|ed.
			++hidden	Open the terminal in a hidden buffer,
					no window will be used.
			++norestore	Do not include this terminal window
					in a session file.
			++kill={how}	When trying to close the terminal
					window kill the job with {how}.  See
					|term_setkill()| for the values.
			++rows={height} Use {height} for the terminal window
					height.  If the terminal uses the full
					Vim height (no window above or below
					the terminal window) the command line
					height will be reduced as needed.
			++cols={width}  Use {width} for the terminal window
					width. If the terminal uses the full
					Vim width (no window left or right of
					the terminal window) this value is
					ignored.
			++eof={text}	when using [range]: text to send after
					the last line was written. Cannot
					contain white space.  A CR is
					appended.  For MS-Windows the default
					is to send CTRL-D.
					E.g. for a shell use "++eof=exit" and
					for Python "++eof=exit()".  Special
					codes can be used like with `:map`,
					e.g. "<C-Z>" for CTRL-Z.

			If you want to use more options use the |term_start()|
			function.
			If you want to split the window vertically, use: >
				:vertical terminal
<			Or short: >
				:vert ter

When the buffer associated with the terminal is forcibly unloaded or wiped out
the job is killed, similar to calling `job_stop(job, "kill")` .
Closing the window normally results in |E947|.  When a kill method was set
with "++kill={how}" or |term_setkill()| then closing the window will use that
way to kill or interrupt the job.  For example: >
	:term ++kill=term tail -f /tmp/log

So long as the job is running the window behaves like it contains a modified
buffer.  Trying to close the window with `CTRL-W :quit` fails.  When using
`CTRL-W :quit!` the job is ended.  The text in the window is lost.  The buffer
still exists, but getting it in a window with `:buffer` will show an empty
buffer.

Trying to close the window with `CTRL-W :close` also fails.   Using
`CTRL-W :close!` will close the window and make the buffer hidden.

You can use `CTRL-W :hide` to close the terminal window and make the buffer
hidden, the job keeps running.  The `:buffer` command can be used to turn the
current window into a terminal window.  If there are unsaved changes this
fails, use ! to force, as usual.

To have a background job run without a window, and open the window when it's
done, use options like this: >
	:term ++hidden ++open make
Note that the window will open at an unexpected moment, this will interrupt
what you are doing.

							*E947* *E948*
So long as the job is running, the buffer is considered modified and Vim
cannot be quit easily, see |abandon|.

When the job has finished and no changes were made to the buffer: closing the
window will wipe out the buffer.

Before changes can be made to a terminal buffer, the 'modifiable' option must
be set.  This is only possible when the job has finished.  At the first change
the buffer will become a normal buffer and the highlighting is removed.
You may want to change the buffer name with |:file| to be able to write, since
the buffer name will still be set to the command.


Resizing ~
							*terminal-resizing*
The size of the terminal can be in one of three modes:

1. The 'termwinsize' option is empty: The terminal size follows the window
   size.  The minimal size is 2 screen lines with 10 cells.

2. The 'termwinsize' option is "rows*cols", where "rows" is the minimal number
   of screen rows and "cols" is the minimal number of cells.

3. The 'termwinsize' option is "rowsXcols" (where the x is upper or lower
   case).  The terminal size is fixed to the specified number of screen lines
   and cells.  If the window is bigger there will be unused empty space.

If the window is smaller than the terminal size, only part of the terminal can
be seen (the lower-left part).

The |term_getsize()| function can be used to get the current size of the
terminal.  |term_setsize()| can be used only when in the first or second mode,
not when 'termwinsize' is "rowsXcols".


Terminal-Job and Terminal-Normal mode ~
						*Terminal-mode* *Terminal-Job*
When the job is running the contents of the terminal is under control of the
job.  That includes the cursor position.  Typed keys are sent to the job.
The terminal contents can change at any time.  This is called Terminal-Job
mode.

Use CTRL-W N (or 'termwinkey' N) to switch to Terminal-Normal mode.  Now the
contents of the terminal window is under control of Vim, the job output is
suspended.  CTRL-\ CTRL-N does the same.

Terminal-Job mode is where |:tmap| mappings are applied. Keys sent by
|term_sendkeys()| are not subject to tmap, but keys from |feedkeys()| are.

It is not possible to enter Insert mode from Terminal-Job mode.

						*Terminal-Normal* *E946*
In Terminal-Normal mode you can move the cursor around with the usual Vim
commands, Visually mark text, yank text, etc.  But you cannot change the
contents of the buffer.  The commands that would start insert mode, such as
'i' and 'a', return to Terminal-Job mode.  The window will be updated to show
the contents of the terminal. |:startinsert| is ineffective.

In Terminal-Normal mode the statusline and window title show "(Terminal)".  If
the job ends while in Terminal-Normal mode this changes to
"(Terminal-finished)".

When the job outputs lines in the terminal, such that the contents scrolls off
the top, those lines are remembered and can be seen in Terminal-Normal mode.
The number of lines is limited by the 'termwinscroll' option. When going over
this limit, the first 10% of the scrolled lines are deleted and are lost.


Cursor style ~
							*terminal-cursor-style*
By default the cursor in the terminal window uses a not blinking block.  The
normal xterm escape sequences can be used to change the blinking state and the
shape.  Once focus leaves the terminal window Vim will restore the original
cursor.

An exception is when xterm is started with the "-bc" argument, or another way
that causes the cursor to blink.  This actually means that the blinking flag
is inverted.  Since Vim cannot detect this, the terminal window cursor
blinking will also be inverted.


Session ~
							*terminal-session*
A terminal window will be restored when using a session file, if possible and
wanted.

If "terminal" was removed from 'sessionoptions' then no terminal windows will
be restored.

If the job in the terminal was finished the window will not be restored.

If the terminal can be restored, the command that was used to open it will be
used again.  To change this use the |term_setrestore()| function.  This can
also be used to not restore a specific terminal by setting the command to
"NONE".


Special keys ~
							*terminal-special-keys*
Since the terminal emulator simulates an xterm, only escape sequences that
both Vim and xterm recognize will be available in the terminal window.  If you
want to pass on other escape sequences to the job running in the terminal you
need to set up forwarding.  Example: >
	tmap <expr> <Esc>]b SendToTerm("\<Esc>]b")
	func SendToTerm(what)
	  call term_sendkeys('', a:what)
	  return ''
	endfunc


Unix ~
							*terminal-unix*
On Unix a pty is used to make it possible to run all kinds of commands.  You
can even run Vim in the terminal!  That's used for debugging, see below.

Environment variables are used to pass information to the running job:
    TERM		the name of the terminal, from the 'term' option or
			$TERM in the GUI; falls back to "xterm" if it does not
			start with "xterm"
    ROWS		number of rows in the terminal initially
    LINES		same as ROWS
    COLUMNS		number of columns in the terminal initially
    COLORS		number of colors, 't_Co' (256*256*256 in the GUI)
    VIM_SERVERNAME	v:servername
    VIM_TERMINAL	v:version


MS-Windows ~
							*terminal-ms-windows*
On MS-Windows winpty is used to make it possible to run all kind of commands.
Obviously, they must be commands that run in a terminal, not open their own
window.

You need the following two files from winpty:

    winpty.dll
    winpty-agent.exe

You can download them from the following page:

    https://github.com/rprichard/winpty

Just put the files somewhere in your PATH.  You can set the 'winptydll' option
to point to the right file, if needed.  If you have both the 32-bit and 64-bit
version, rename to winpty32.dll and winpty64.dll to match the way Vim was
build.

Environment variables are used to pass information to the running job:
    VIM_SERVERNAME	v:servername

==============================================================================
2. Terminal communication			 *terminal-communication*

There are several ways to communicate with the job running in a terminal:
- Use |term_sendkeys()| to send text and escape sequences from Vim to the job.
- Use the JSON API to send encoded commands from the job to Vim.
- Use the |client-server| mechanism. This works on machines with an X server
  and on MS-Windows.


Vim to job: term_sendkeys() ~
							*terminal-to-job*
This allows for remote controlling the job running in the terminal.  It is a
one-way mechanism.  The job can update the display to signal back to Vim.
For example, if a shell is running in a terminal, you can do: >
	call term_sendkeys(buf, "ls *.java\<CR>")

This requires for the job to be in the right state where it will do the right
thing when receiving the keys.  For the above example, the shell must be
waiting for a command to be typed.

For a job that was written for the purpose, you can use the JSON API escape
sequence in the other direction.  E.g.: >
	call term_sendkeys(buf, "\<Esc>]51;["response"]\x07")


Job to Vim: JSON API ~
							*terminal-api*
The job can send JSON to Vim, using a special escape sequence.  The JSON
encodes a command that Vim understands.  Example of such a message: >
	<Esc>]51;["drop", "README.md"]<07>

The body is always a list, making it easy to find the end: ]<07>.
The <Esc>]51;msg<07> sequence is reserved by xterm for "Emacs shell", which is
similar to what we are doing here.

Currently supported commands:

	call {funcname} {argument}

		Call a user defined function with {argument}.
		The function is called with two arguments: the buffer number
		of the terminal and {argument}, the decoded JSON argument. 
		The function name must start with "Tapi_" to avoid
		accidentally calling a function not meant to be used for the
		terminal API.
		The user function should sanity check the argument.
		The function can use |term_sendkeys()| to send back a reply.
		Example in JSON: >
			["call", "Tapi_Impression", ["play", 14]]
<		Calls a function defined like this: >
			function Tapi_Impression(bufnum, arglist)
			  if len(a:arglist) == 2
			    echomsg "impression " . a:arglist[0]
			    echomsg "count " . a:arglist[1]
			  endif
			endfunc
<		Output from `:echo` may be erased by a redraw, use `:echomsg`
		to be able to see it with `:messages`.

	drop {filename} [options]

		Let Vim open a file, like the `:drop` command.  If {filename}
		is already open in a window, switch to that window.  Otherwise
		open a new window to edit {filename}.
		Note that both the job and Vim may change the current
		directory, thus it's best to use the full path.

		[options] is only used when opening a new window.  If present,
		it must be a Dict.  Similarly to |++opt|, These entries are recognized:
		  "ff"		file format: "dos", "mac" or "unix"
		  "fileformat"	idem
		  "enc"		overrides 'fileencoding'
		  "encoding"	idem
		  "bin"		sets 'binary'
		  "binary"	idem
		  "nobin"	resets 'binary'
		  "nobinary"	idem
		  "bad"		specifies behavior for bad characters, see
				|++bad|

		Example in JSON: >
			["drop", "path/file.txt", {"ff": "dos"}]

A trick to have Vim send this escape sequence: >
	exe "set t_ts=\<Esc>]51; t_fs=\x07"
	let &titlestring = '["call","Tapi_TryThis",["hello",123]]'
	redraw
	set t_ts& t_fs&

Rationale: Why not allow for any command or expression?  Because that might
create a security problem.


Using the client-server feature ~
						*terminal-client-server*
This only works when v:servername is not empty.  If needed you can set it,
before opening the terminal, with: >
	call remote_startserver('vim-server')

$VIM_SERVERNAME is set in the terminal to pass on the server name.

In the job you can then do something like: >
	vim --servername $VIM_SERVERNAME --remote +123 some_file.c
This will open the file "some_file.c" and put the cursor on line 123.

==============================================================================
3. Remote testing					*terminal-testing*

Most Vim tests execute a script inside Vim.  For some tests this does not
work, running the test interferes with the code being tested.  To avoid this
Vim is executed in a terminal window.  The test sends keystrokes to it and
inspects the resulting screen state.

Functions ~

|term_sendkeys()|	send keystrokes to a terminal (not subject to tmap)
|term_wait()|		wait for screen to be updated
|term_scrape()|		inspect terminal screen


==============================================================================
4. Diffing screen dumps					*terminal-diff*

In some cases it can be bothersome to test that Vim displays the right
characters on the screen.  E.g. with syntax highlighting.  To make this
simpler it is possible to take a screen dump of a terminal and compare it to
an expected screen dump.

Vim uses the window size, text, color and other attributes as displayed.  The
Vim screen size, font and other properties do not matter.  Therefore this
mechanism is portable across systems.  A conventional screenshot would reflect
all differences, including font size and family.


Writing a screen dump test for Vim ~
							*terminal-dumptest*
For an example see the Test_syntax_c() function in
src/testdir/test_syntax.vim.  The main parts are:
- Write a file you want to test with. This is useful for testing syntax
  highlighting.  You can also start Vim with en empty buffer.
- Run Vim in a terminal with a specific size.  The default is 20 lines of 75
  characters.  This makes sure the dump is always this size.  The function
  RunVimInTerminal() takes care of this.  Pass it the arguments for the Vim
  command.
- Send any commands to Vim using |term_sendkeys()|.  For example: >
	call term_sendkeys(buf, ":echo &lines &columns\<CR>")
- Check that the screen is now in the expected state, using
  VerifyScreenDump().  This expects the reference screen dump to be in the
  src/testdir/dumps/ directory.  Pass the name without ".dump".  It is
  recommended to use the name of the test function and a sequence number, so
  that we know what test is using the file.
- Repeat sending commands and checking the state.
- Finally stop Vim by calling StopVimInTerminal().

The first time you do this you won't have a screen dump yet.  Create an empty
file for now, e.g.: >
	touch src/testdir/dumps/Test_function_name_01.dump

The test will then fail, giving you the command to compare the reference dump
and the failed dump, e.g.: >
	call term_dumpdiff("Test_func.dump.failed", "dumps/Test_func.dump")

Use this command in Vim, with the current directory set to src/testdir.
Once you are satisfied with the test, move the failed dump in place of the
reference: >
	:!mv Test_func.dump.failed dumps/Test_func.dump


Creating a screen dump ~
							*terminal-screendump*

To create the screen dump, run Vim (or any other program) in a terminal and
make it show the desired state.  Then use the |term_dumpwrite()| function to
create a screen dump file.  For example: >
	:call term_dumpwrite(77, "mysyntax.dump")

Here "77" is the buffer number of the terminal.  Use `:ls!` to see it.

You can view the screen dump with |term_dumpload()|: >
	:call term_dumpload("mysyntax.dump")

To verify that Vim still shows exactly the same screen, run Vim again with
exactly the same way to show the desired state.  Then create a screen dump
again, using a different file name: >
	:call term_dumpwrite(88, "test.dump")

To assert that the files are exactly the same use |assert_equalfile()|: >
	call assert_equalfile("mysyntax.dump", "test.dump")

If there are differences then v:errors will contain the error message.


Comparing screen dumps ~
						*terminal-diffscreendump*

|assert_equalfile()| does not make it easy to see what is different.
To spot the problem use |term_dumpdiff()|: >
	call term_dumpdiff("mysyntax.dump", "test.dump")

This will open a window consisting of three parts:
1.  The contents of the first dump
2.  The difference between the first and second dump
3.  The contents of the second dump

You can usually see what differs in the second part.  Use the 'ruler' to
relate it to the position in the first or second dump.  Letters indicate the
kind of difference:
	X	different character
	>	cursor in first but not in second
	<	cursor in second but not in first
	w	character width differs (single vs double width)
	f	foreground color differs
	b	background color differs
	a	attribute differs (bold, underline, reverse, etc.)
	?	character missing in both
	+	character missing in first
	-	character missing in second

Alternatively, press "s" to swap the first and second dump. Do this several
times so that you can spot the difference in the context of the text.

==============================================================================
5. Debugging				*terminal-debug* *terminal-debugger*

The Terminal debugging plugin can be used to debug a program with gdb and view
the source code in a Vim window.  Since this is completely contained inside
Vim this also works remotely over an ssh connection.

When the |+terminal| feature is missing, the plugin will use the "prompt"
buffer type, if possible.  The running program will then use a newly opened
terminal window.  See |termdebug-prompt| below for details.


Starting ~
							*termdebug-starting*
Load the plugin with this command: >
	packadd termdebug
<							*:Termdebug*
To start debugging use `:Termdebug` or `:TermdebugCommand` followed by the
command name, for example: >
	:Termdebug vim

This opens two windows:

gdb window	A terminal window in which "gdb vim" is executed.  Here you
		can directly interact with gdb.  The buffer name is "!gdb".

program window	A terminal window for the executed program.  When "run" is
		used in gdb the program I/O will happen in this window, so
		that it does not interfere with controlling gdb.  The buffer
		name is "gdb program".

The current window is used to show the source code.  When gdb pauses the
source file location will be displayed, if possible.  A sign is used to
highlight the current position, using highlight group debugPC.	 

If the buffer in the current window is modified, another window will be opened
to display the current gdb position.  You can use `:Winbar` to add a window
toolbar there.

Focus the terminal of the executed program to interact with it.  This works
the same as any command running in a terminal window.

When the debugger ends, typically by typing "quit" in the gdb window, the two
opened windows are closed.

Only one debugger can be active at a time.
							*:TermdebugCommand*
If you want to give specific commands to the command being debugged, you can
use the `:TermdebugCommand` command followed by the command name and
additional parameters. >
	:TermdebugCommand vim --clean -c ':set nu'

Both the `:Termdebug` and `:TermdebugCommand` support an optional "!" bang
argument to start the command right away, without pausing at the gdb window
(and cursor will be in the debugged window).  For example: >
	:TermdebugCommand! vim --clean

To attach gdb to an already running executable or use a core file, pass extra
arguments.  E.g.: >
	:Termdebug vim core
	:Termdebug vim 98343

If no argument is given, you'll end up in a gdb window, in which you need to
specify which command to run using e.g. the gdb `file` command.


Example session ~
							*termdebug-example*
Start in the Vim "src" directory and build Vim: >
	% make
Start Vim: >
	% ./vim
Load the termdebug plugin and start debugging Vim: >
	:packadd termdebug
	:Termdebug vim
You should now have three windows:
    source  - where you started, has a window toolbar with buttons
    gdb	    - you can type gdb commands here
    program - the executed program will use this window

You can use CTRL-W CTRL-W or the mouse to move focus between windows.
Put focus on the gdb window and type: >
	break ex_help
	run
Vim will start running in the program window. Put focus there and type: >
	:help gui
Gdb will run into the ex_help breakpoint.  The source window now shows the 
ex_cmds.c file.  A red "1 " marker will appear in the signcolumn where the
breakpoint was set.  The line where the debugger stopped is highlighted.  You
can now step through the program.  Let's use the mouse: click on the "Next"
button in the window toolbar.  You will see the highlighting move as the
debugger executes a line of source code.

Click "Next" a few times until the for loop is highlighted.  Put the cursor on
the end of "eap->arg", then click "Eval" in the toolbar.  You will see this
displayed:
	"eap->arg": 0x555555e68855 "gui" ~
This way you can inspect the value of local variables.  You can also focus the
gdb window and use a "print" command, e.g.: >
	print *eap
If mouse pointer movements are working, Vim will also show a balloon when the
mouse rests on text that can be evaluated by gdb.

Now go back to the source window and put the cursor on the first line after
the for loop, then type: >
	:Break
You will see a ">>" marker appear, this indicates the new breakpoint.  Now
click "Cont" in the toolbar and the code until the breakpoint will be
executed.

You can type more advanced commands in the gdb window.  For example, type: >
	watch curbuf
Now click "Cont" in the toolbar (or type "cont" in the gdb window). Execution
will now continue until the value of "curbuf" changes, which is in do_ecmd().
To remove this watchpoint again type in the gdb window: >
	delete 3

You can see the stack by typing in the gdb window: >
	where
Move through the stack frames, e.g. with: >
	frame 3
The source window will show the code, at the point where the call was made to
a deeper level.


Stepping through code ~
							*termdebug-stepping*
Put focus on the gdb window to type commands there.  Some common ones are:
- CTRL-C	interrupt the program
- next		execute the current line and stop at the next line
- step		execute the current line and stop at the next statement,
		entering functions
- finish	execute until leaving the current function
- where		show the stack
- frame N	go to the Nth stack frame
- continue	continue execution

						*:Run* *:Arguments*
In the window showing the source code these commands can be used to control
gdb:
 `:Run` [args]	    run the program with [args] or the previous arguments
 `:Arguments` {args}  set arguments for the next `:Run`

 *:Break*	set a breakpoint at the current line; a sign will be displayed
 *:Clear*	delete the breakpoint at the current line

 *:Step*	execute the gdb "step" command
 *:Over*	execute the gdb "next" command (`:Next` is a Vim command)
 *:Finish*	execute the gdb "finish" command
 *:Continue*	execute the gdb "continue" command
 *:Stop*	interrupt the program

If 'mouse' is set the plugin adds a window toolbar with these entries:
  Step		`:Step`
  Next		`:Over`
  Finish	`:Finish`
  Cont		`:Continue`
  Stop		`:Stop`
  Eval		`:Evaluate`
This way you can use the mouse to perform the most common commands.  You need
to have the 'mouse' option set to enable mouse clicks.
								*:Winbar*
You can add the window toolbar in other windows you open with: >
  :Winbar

If gdb stops at a source line and there is no window currently showing the
source code, a new window will be created for the source code.  This also
happens if the buffer in the source code window has been modified and can't be
abandoned.

Gdb gives each breakpoint a number.  In Vim the number shows up in the sign
column, with a red background.  You can use these gdb commands:
- info break	list breakpoints
- delete N	delete breakpoint N
You can also use the `:Clear` command if the cursor is in the line with the
breakpoint, or use the "Clear breakpoint" right-click menu entry.


Inspecting variables ~
					*termdebug-variables* *:Evaluate*
 `:Evaluate`	    evaluate the expression under the cursor
 `K`		    same
 `:Evaluate` {expr}   evaluate {expr}
 `:'<,'>Evaluate`     evaluate the Visually selected text

This is similar to using "print" in the gdb window.
You can usually shorten `:Evaluate` to `:Ev`.


Other commands ~
							*termdebug-commands*
 *:Gdb*	     jump to the gdb window
 *:Program*    jump to the window with the running program
 *:Source*     jump to the window with the source code, create it if there
	     isn't one


Prompt mode ~
						*termdebug-prompt*
When the |+terminal| feature is not supported and on MS-Windows, gdb will run
in a buffer with 'buftype' set to "prompt".  This works slightly differently:
- The gdb window will be in Insert mode while typing commands.  Go to Normal
  mode with <Esc>, then you can move around in the buffer, copy/paste, etc.
  Go back to editing the gdb command with any command that starts Insert mode,
  such as `a` or `i`.
- The program being debugged will run in a separate window.  On MS-Windows
  this is a new console window.  On Unix, if the |+terminal| feature is
  available a Terminal window will be opened to run the debugged program in.

						*termdebug_use_prompt*
Prompt mode can be used even when the |+terminal| feature is present with: >
	let g:termdebug_use_prompt = 1


Communication ~
						*termdebug-communication*
There is another, hidden, buffer, which is used for Vim to communicate with
gdb.  The buffer name is "gdb communication".  Do not delete this buffer, it
will break the debugger.

Gdb has some weird behavior, the plugin does its best to work around that.
For example, after typing "continue" in the gdb window a CTRL-C can be used to
interrupt the running program.  But after using the MI command
"-exec-continue"  pressing CTRL-C does not interrupt.  Therefore you will see
"continue" being used for the `:Continue` command, instead of using the
communication channel.


Customizing ~

GDB command						 *termdebug-customizing*

To change the name of the gdb command, set the "termdebugger" variable before
invoking `:Termdebug`: >
	let termdebugger = "mygdb"
<							*gdb-version*
Only debuggers fully compatible with gdb will work.  Vim uses the GDB/MI
interface.  The "new-ui" command  requires gdb version 7.12 or later.  if you
get this error:
	Undefined command: "new-ui". Try "help".~
Then your gdb is too old.


Colors						*hl-debugPC* *hl-debugBreakpoint*

The color of the signs can be adjusted with these highlight groups:
- debugPC		the current position
- debugBreakpoint	a breakpoint

The defaults are, when 'background' is "light":
  hi debugPC term=reverse ctermbg=lightblue guibg=lightblue
  hi debugBreakpoint term=reverse ctermbg=red guibg=red

When 'background' is "dark":
  hi debugPC term=reverse ctermbg=darkblue guibg=darkblue
  hi debugBreakpoint term=reverse ctermbg=red guibg=red


Shorcuts						*termdebug_shortcuts*

You can define your own shortcuts (mappings) to control gdb, that can work in
any window, using the TermDebugSendCommand() function.  Example: >
	map ,w :call TermDebugSendCommand('where')<CR>
The argument is the gdb command.


Popup menu						*termdebug_popup*

By default the Termdebug plugin sets 'mousemodel' to "popup_setpos" and adds
these entries to the popup menu:
	Set breakpoint		`:Break`
	Clear breakpoint	`:Clear`
	Evaluate		`:Evaluate`
If you don't want this then disable it with: >
	let g:termdebug_popup = 0


Vim window width						*termdebug_wide*

To change the width of the Vim window when debugging starts, and use a
vertical split: >
  let g:termdebug_wide = 163
This will set &columns to 163 when `:Termdebug` is used.  The value is restored
when quitting the debugger.
If g:termdebug_wide is set and &columns is already larger than
g:termdebug_wide then a vertical split will be used without changing &columns.
Set it to 1 to get a vertical split without every changing &columns (useful
for when the terminal can't be resized by Vim).



 vim:tw=78:ts=8:noet:ft=help:norl: