view etc/TERMS @ 788:026c5bf9c134

[xemacs-hg @ 2002-03-21 07:29:57 by ben] chartab.c: Fix bugs in implementation and doc strings. config.h.in: Add foo_checking_assert_at_line() macros. Not clear whether these are actually useful, though; I'll take them out if not. symsinit.h, emacs.c: Some improvements to the timeline. Rearrange a bit the init calls. Add call for reinit_vars_of_object_mswindows() and declare in symsinit.h. event-Xt.c, event-gtk.c, event-msw.c, event-stream.c, event-tty.c, events.c, events.h: Introduce new event methods for printing, comparing, and hashing magic events, to avoid event-type-specific stuff that had crept into events.c. (And was crashing, since the channel in MS Windows magic events may be nil.) Implement the methods in event-{tty,gtk,Xt,mswindows}.c. Make wrapping functions event_stream_{compare,hash,format}_magic_event() to check if everything's OK and call the actual callback. Fix events.c to use the new methods. Add a new event-stream-operation EVENT_STREAM_NOTHING -- event stream not actually required to be able to do anything, just be open. (#### This event-stream-operation stuff needs to be rethought.) Fixed describe_event() in event-Xt.c to print its output to a stream, not always to stderr, so it can be used elsewhere. (e.g. in print-event when a magic event is encountered?) lisp.h, lrecord.h: Define new assert_at_line(), for use in asserts inside of inline functions. The assert will report the line and file of the inline function, which is almost certainly not what you want as it's useless. what you want to see is where the pseudo-macro was called from. So, when error-checking is on, we pass in the line and file into the macros, for accurate printout using assert_at_line(). Happens only when error-checking is defined so doesn't slow down non-error-checking builds. Fix XCHAR, XINT, XCHAR_OR_INT, XFOO, and wrap_foo() in this fashion. lstream.c, lstream.h: Add resizing_buffer_to_lisp_string(). objects-gtk.c: Fix typo. objects-msw.c: Implement a smarter way of determining whether a font matches a charset. Formerly we just looked at the "script" element of the font spec, converted it to a code page, and compared it with the code page derived from the charset. Now, as well as doing this, we ask the font for the list of unicode ranges it supports, see what range the charset falls into (#### bogus! need to do this char-by-char), and see if any of the font's supported ranges include the charset's range. also do some caching in Vfont_signature_data of previous inquiries. charset.h, text.c, mule-charset.c: New fun; extracted out of Fmake_char() and declare prototype in charset.h. text.h: introduce assert_by_line() to make REP_BYTES_BY_FIRST_BYTE report the file and line more accurately in an assertion failure. unicode.c: make non-static (used in objects-msw.c), declare in charset.h. mule\mule-category.el: Start implementing a category API compatible with FSF. Not there yet. We need improvements to char-tables. mule\mule-charset.el: Copy translation table code from FSF 21.1 and fix up. Eventually we'll have them in XEmacs. (used in ccl) Not here quite yet, and we need some improvements to char-tables. mule\cyril-util.el, mule\cyrillic.el, mule\devan-util.el, mule\ethio-util.el, mule\korea-util.el, mule\mule-tty-init.el, mule\tibet-util.el, mule\viet-util.el, mule\vietnamese.el: Fix numerous compilation warnings. Fix up code related to translation tables and other types of char-tables. menubar-items.el: Move the frame commands from the View menu to the File menu, to be consistent with how most other programs do things. Move less-used revert/recover items to a submenu. Make "recover" not prompt for a file, but recover the current buffer. TODO.ben-mule-21-5: Create bug list for latest problems.
author ben
date Thu, 21 Mar 2002 07:31:30 +0000
parents 376386a54a3c
children
line wrap: on
line source

This file describes what you must or might want to do to termcap entries
to make terminals work properly and efficiently with Emacs.  Information
on likely problems with specific types of terminals appears at the end
of the file.

*** What you want in a terminal ***

Vital
1. Easy to compute suitable padding for.
2. Never ever sends ^S/^Q unless you type them, at least in one mode.

Nice for speed
1. Supports insert/delete of multiple lines in one command.
2. Same for multiple characters, though doing them one by
one is usually fast enough except on emulators running on
machines with bitmap screens.

Nice for usability
1. Considerably more than 24 lines.
2. Meta key (shift-like key that controls the 0200 bit
in every character you type).

*** New termcap strings ***

Emacs supports certain termcap strings that are not described in the
4.2 manual but appear to be standard in system V.  The one exception
is `cS', which I invented.

`AL'    insert several lines.  Takes one parameter, the number of
        lines to be inserted.  You specify how to send this parameter
	using a %-construct, just like the cursor positions in the `cm'
	string.

`DL'    delete several lines.  One parameter.

`IC'    insert several characters.  One parameter.

`DC'    delete several characters.  One parameter.

`rp'    repeat a character.  Takes two parameters, the character
        to be repeated and the number of times to repeat it.
	Most likely you will use `%.' for sending the character
	to be repeated.  Emacs interprets a padding spec with a *
	as giving the amount of padding per repetition.

	WARNING: Many terminals have a command to repeat the
	*last character output* N times.  This means that the character
	will appear N+1 times in a row when the command argument is N.
	However, the `rp' string's parameter is the total number of
	times wanted, not one less.  Therefore, such repeat commands
	may be used in an `rp' string only if you use Emacs's special
	termcap operator `%a-c\001' to subtract 1 from the repeat count
	before substituting it into the string.  It is probably safe
	to use this even though the Unix termcap does not accept it
	because programs other than Emacs probably won't look for `rp'
	anyway.
	
`cs'    set scroll region.  Takes two parameters, the vertical
	positions of the first line to include in the scroll region
	and the last line to include in the scroll region.
	Both parameters are origin-zero.  The effect of this
	should be to cause a following insert-line or delete-line
	not to move lines below the bottom of the scroll region.

	This is not the same convention that Emacs version 16 used.
	That is because I was led astray by unclear documentation
	of the meaning of %i in termcap strings.  Since the termcap
	documentation for `cs' is also unclear, I had to deduce the
	correct parameter conventions from what would make the VT-100's
	`cs' string work properly.  From an incorrect assumption about
	%i, I reached an incorrect conclusion about `cs', but the result
	worked correctly on the VT100 and ANSII terminals.  In Emacs
	version 17, both `cs' and %i work correctly.

	The version 16 convention was to pass, for the second parameter,
	the line number of the first line beyond the end of the
	scroll region.

`cS'    set scroll region.  Differs from `cs' in taking parameters
	differently.  There are four parameters:
	1. Total number of lines on the screen.
	2. Number of lines above desired scroll region.
	3. Number of lines below (outside of) desired scroll region.
	4. Total number of lines on the screen, like #1.
	This is because an Ambassador needs the parameters like this.

`cr', `do', `le'
	Emacs will not attempt to use ^M, ^J or ^H for cursor motion
	unless these capabilities are present and say to use those
	characters.

`km'    Says the terminal has a Meta key.

Defining these strings is important for getting maximum performance
from your terminal.

Make sure that the `ti' string sets all modes needed for editing
in Emacs.  For example, if your terminal has a mode that controls
wrap at the end of the line, you must decide whether to specify
the `am' flag in the termcap entry; whichever you decide, the `ti'
string should contain commands to set the mode that way.
(Emacs also sends the `vs' string after the `ti' string.
You can put the mode-setting commands in either one of them.)

*** Specific Terminal Types ***

Watch out for termcap entries for Ann Arbor Ambassadors that
give too little padding for clear-screen.  7.2 msec per line is right.
These are the strings whose padding you probably should change:
    :al=1*\E[L:dl=1*\E[M:cd=7.2*\E[J:cl=7.2*\E[H\E[J:
I have sometimes seen `\E[2J' at the front of the `ti' string;
this is a clear-screen, very slow, and it can cause you to get
Control-s sent by the terminal at startup.  I recommend removing
the `\E[2J' from the `ti' string.
The `ti' or `vs' strings also usually need stuff added to them, such as
    \E[>33;52;54h\E[>30;37;38;39l
You might want to add the following to the `te' or `ve' strings:
    \E[>52l\E[>37h
The following additional capabilities will improve performance:
    :AL=1*\E[%dL:DL=1*\E[%dM:IC=4\E[%d@:DC=4\E[%dP:rp=1*%.\E[%a-c\001%db:
If you find that the Meta key does not work, make sure that
    :km:
is present in the termcap entry.

Watch out for termcap entries for VT100's that fail to specify
the `sf' string, or that omit the padding needed for the `sf' and `sr'
strings (2msec per line affected).  What you need is
    :sf=2*^J:sr=2*\EM:cs=\E[%i%d;%dr:

The Concept-100 and Concept-108 have many modes that `ti' strings
often fail to initialize.  If you have problems on one of these
terminals, that is probably the place to fix them.  These terminals
can support an `rp' string.

Watch out on HP terminals for problems with standout disappearing on
part of the mode line.  These problems are due to the absence of
:sg#0: which some HP terminals need.

The vi55 is said to require `ip=2'.

The Sun console should have these capabilities for good performance.
	   :AL=\E[%dL:DL=\E[%dM:IC=\E[%d@:DC=\E[%dP:

The vt220 needs to be set to vt220 mode, 7 bit, space parity
in order to work fully with TERM=vt220.

If you are using a LAT terminal concentrator, you need to issue these
commands to turn off flow control:

    set port flow control disable
    define port flow control disable

On System V, in the terminfo database, various terminals may have
the `xt' flag that should not have it.  `xt' should be present only
for the Teleray 1061 or equivalent terminal.

In particular, System V for the 386 often has `xt' for terminal type
AT386 or AT386-M, which is used for the console.  You should delete
this flag.  Here is how:

You can get a copy of the terminfo "source" for at386 using the
command: `infocmp at386 >at386.tic'.  Edit the file at386.tic and remove
the `xt' flag.  Then compile the new entry with: `tic at386.tic'.

It is also reported that these terminal types sometimes have the wrong
reverse-scroll string.  It should be \E[T, but sometimes is given as \E[S.

Here is what watserv1!maytag!focsys!larry recommends for these terminals:

# This copy of the terminfo description has been fixed.
# The suggestions came from a number of usenet postings.
#
# Intel AT/386 for color card with monochrome display
#
AT386-M|at386-m|386AT-M|386at-m|at/386 console,
	am, bw, eo, xon,
	cols#80, lines#25,
	acsc=``a1fxgqh0jYk?lZm@nEooppqDrrsstCu4vAwBx3yyzz{{||}}~~,
	bel=^G, blink=\E[5m, bold=\E[1m, cbt=\E[Z,
	clear=\E[2J\E[H,
	cr=\r, cub=\E[%p1%dD, cub1=\E[D, cud=\E[%p1%dB,
	cud1=\E[B, cuf=\E[%p1%dC, cuf1=\E[C,
	cup=\E[%i%p1%02d;%p2%02dH, cuu=\E[%p1%dA, cuu1=\E[A,
	dch=\E[%p1%dP, dch1=\E[P, dl=\E[%p1%dM, dl1=\E[1M,
	ech=\E[%p1%dX,ed=\E[J, el=\E[K, el1=\E[1K\E[X, flash=^G, home=\E[H,
	hpa=\E[%i%p1%dG, ich=\E[%p1%d@, ich1=\E[1@, il=\E[%p1%dL, il1=\E[1L,
	ind=\E[S, indn=\E[%p1%dS, invis=\E[9m,
	is2=\E[0;10;38m, kbs=\b, kcbt=^], kclr=\E[2J,
	kcub1=\E[D, kcud1=\E[B, kcuf1=\E[C, kcuu1=\E[A,
	kdch1=\E[P, kend=\E[Y, kf1=\EOP, kf10=\EOY, kf11=\EOZ,
	kf12=\EOA, kf2=\EOQ, kf3=\EOR, kf4=\EOS, kf5=\EOT,
	kf6=\EOU, kf7=\EOV, kf8=\EOW, kf9=\EOX, khome=\E[H,
	kich1=\E[@, knp=\E[U, kpp=\E[V, krmir=\E0, rev=\E[7m, ri=\E[T,
	rin=\E[%p1%dT, rmacs=\E[10m, rmso=\E[m, rmul=\E[m,
	sgr=\E[10m\E[0%?%p1%p3%|%t;7%;%?%p2%t;4%;%?%p4%t;5%;%?%p6%t;1%;%?%p9%t;12%;%?%p7%t;9%;m,
	sgr0=\E[0;10m, smacs=\E[12m, smso=\E[7m, smul=\E[4m,

#
# AT&T 386 color console 
#
AT386|at386|386AT|386at|at/386 console,
	colors#8, ncv#3, pairs#64,
	is2=\E[0;10;39m,
	op=\E[0m, 
	setb=\E[%?%p1%{0}%=%t40m
                %e%p1%{1}%=%t44m
                %e%p1%{2}%=%t42m
                %e%p1%{3}%=%t46m
                %e%p1%{4}%=%t41m
                %e%p1%{5}%=%t45m
                %e%p1%{6}%=%t43m
                %e%p1%{7}%=%t47m%;,
	setf=\E[%?%p1%{0}%=%t30m
                %e%p1%{1}%=%t34m
                %e%p1%{2}%=%t32m
                %e%p1%{3}%=%t36m
                %e%p1%{4}%=%t31m
                %e%p1%{5}%=%t35m
                %e%p1%{6}%=%t33m
                %e%p1%{6}%=%t33m
                %e%p1%{7}%=%t37m%;,
	use=at386-m,
#
# Color console version that supports underline but maps blue
# foreground color to cyan.
#
AT386-UL|at386-ul|386AT-UL|386at-ul|at/386 console,
	is2=\E[0;10;38m,
	use=at386,