Joseph Rosevear wrote:
> Hello, friends,
>
> I'm adjusting still to Slackware 12.0. For example, I'm finally using
> the generic kernel instead of the huge one, but...
>
> One thing about 12.0 has me stymied. I ran a test. This doesn't
> happen in my Slackware 10.2 installation:
>
> 1. Make this script:
>
> #!/bin/sh
> #try
>
> echo say something
> read answer
> echo $answer
>
> 2. chmod u+x try
>
> 3. ./try
>
> 4. Try will ask you to "say something". So type a few characters.
> Then try to use the backspace key before you press enter. If you
> are running in xterm in X (I tried fvwm2 and twm), then pressing the
> backspace key produces "^H" instead of actually doing a backspace.
>
> In Slackware 10.2 the same test actually *does* the backspace which is
> what I want. What good is it to get "^H" instead of *doing* the
> backspace?
>
> Note that when not in X both 10.2 and 12.0 *do* the backspace, so this
> seems to be related to X.
>
> Can anyone help me with this? I tried tinkering with xmodmap and
> .screenrc, and I found this article on the web:
>
> http://www.cs.brown.edu/people/dnr/t...e-solution.txt
>
> But I haven't solved the problem yet. I just want backspace to *do*
> the backspace *all* the time whether I'm in X or not.
>
> Oh bother. I just solved the problem. I read the article I referenced
> above thinking "Man, I better read this."
>
> Well, the answer is use
>
> stty erase `tput kbs`
>
> at the command line in xterm first and all uses of script try will
> function in the desired way (with real backspace).
>
> I'm still troubled, though. Why the change (from 10.2 to 12.0) so that
> *now* "stty erase `tput kbs`" is needed to get the desired behavior?
>
> I'll quit now before I reveal any more foolishness.
>
> -Joe

It works with me these X resources:

XTerm.ttyModes: erase ^?
XTerm.vt100.backarrowKey: false

In a bash script you couls also
read -e answer

this way you could use all the functionality of readline (cursor keys,
history, etc.)

Olive

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 2008-02-20, Joseph Rosevear wrote:
> read answer
> echo $answer
> 4. Try will ask you to "say something". So type a few characters.
> Then try to use the backspace key before you press enter. If you
> are running in xterm in X (I tried fvwm2 and twm), then pressing the
> backspace key produces "^H" instead of actually doing a backspace.

This is entirely a terminal emulation problem.

> In Slackware 10.2 the same test actually *does* the backspace which is
> what I want. What good is it to get "^H" instead of *doing* the
> backspace?

In many terminal types, the backspace key sends a CONTOL-H (or
abbreviated, ^H) to terminate the previous character. If you try it,
you may find that works in place of backspace within a lot of apps.[0]

> Note that when not in X both 10.2 and 12.0 *do* the backspace, so this
> seems to be related to X.

Your guess is a good one, but unfortunately incorrect. The terminal
emulation used in the console and in xterm is quite different. These
differences are contained in the terminfo database (found in
/usr/share/terminfo).

> Oh bother. I just solved the problem. I read the article I referenced
> above thinking "Man, I better read this."
>
> Well, the answer is use
>
> stty erase `tput kbs`
>
> at the command line in xterm first and all uses of script try will
> function in the desired way (with real backspace).

You can put this in your .bashrc to handle it for you.

if [ "$TERM" = "xterm" ]; then
stty erase $(tput kbs)
fi

> I'm still troubled, though. Why the change (from 10.2 to 12.0) so that
> *now* "stty erase `tput kbs`" is needed to get the desired behavior?

Probably wasn't intentional. Has to be a change in the terminfo
database. I don't have the ncurses source for 10.2 handy, but I
noticed that between 10.1 and 12.0 an xterm.terminfo seemingly
customized for Slackware has been added in source/l/ncurses. If some
one can verify that this is the culprit, Pat might be willing to accept
a patch to it.

> I'll quit now before I reveal any more foolishness.

On the contrary! It's a good question.

[0] Similarly, ^W backspaces over the last word. What constitutes a
word varies depending on the application. Sometimes only whitespace
constitutes the begining and ending of a word; other times,
non-alphanumeric characters are not considered to be parts of a word.

- --
It is better to hear the rebuke of the wise,
Than for a man to hear the song of fools.
Ecclesiastes 7:5
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)

iD8DBQFHvJO6rZS6hX/gvjoRAsgOAKDbUTEX3ZOxvGttsoB1q/W6yTlf9ACcCkUX
N77gmI0bKSoN1Hecgdn6uWA=
=gOgK
-----END PGP SIGNATURE-----

+Alan Hicks+ wrote:

[snip]

> This is entirely a terminal emulation problem.

I don't know much about terminal emulation, but thanks.

[snip]

> In many terminal types, the backspace key sends a CONTOL-H (or
> abbreviated, ^H) to terminate the previous character. If you try it,
> you may find that works in place of backspace within a lot of apps.[0]

Hmmm. My try script gives the same result in xterm for "^H" and
. Without "stty erase `tput kbs`" both print "^H" to the
screen. With "stty erase `tput kbs`" both do a backspace. Is this
behavior an example of what you just described?

However, when not in X (when at the console) "^H" doesn't give the same
result as .

I'm trying to understand. Seems like...

vvv
....there are different levels to this problem. The backspace key and ^H
are two different "key strokes" that have (probably) two different
"codes" associated (mapped). I understand I can change that mapping
(in X) by using

xmodmap -e "keycode blah = something"

where blah is the number of the key and something is the symbol for the
"code".

This "code" then can be interpreted in different ways depending on the
terminal emulation. The terminal emulation begins with terminfo, but
can be changed with the stty command.
^^^

Did I get that right?

> > Note that when not in X both 10.2 and 12.0 *do* the backspace, so this
> > seems to be related to X.

> Your guess is a good one, but unfortunately incorrect. The terminal
> emulation used in the console and in xterm is quite different. These
> differences are contained in the terminfo database (found in
> /usr/share/terminfo).

So, it is not exactly related to X, rather it is related to using xterm
instead of the console?

[snip]

> You can put this in your .bashrc to handle it for you.

> if [ "$TERM" = "xterm" ]; then
> stty erase $(tput kbs)
> fi

Thanks. I did it. Works fine in xterm in 12.0

> > I'm still troubled, though. Why the change (from 10.2 to 12.0) so that
> > *now* "stty erase `tput kbs`" is needed to get the desired behavior?

> Probably wasn't intentional. Has to be a change in the terminfo
> database. I don't have the ncurses source for 10.2 handy, but I
> noticed that between 10.1 and 12.0 an xterm.terminfo seemingly
> customized for Slackware has been added in source/l/ncurses. If some
> one can verify that this is the culprit, Pat might be willing to accept
> a patch to it.

Good idea. The crooked way should sometimes be made straight.

[snip]

> [0] Similarly, ^W backspaces over the last word. What constitutes a
> word varies depending on the application. Sometimes only whitespace
> constitutes the begining and ending of a word; other times,
> non-alphanumeric characters are not considered to be parts of a word.

Interesting.

[skip]

Thanks for you inputs.

-Joe

>> This is entirely a terminal emulation problem.
>
> I don't know much about terminal emulation, but thanks.

Some term progams can be fixed by setting them to send "delete" instead of
"backspace" (ctrl H). I have a term program I use all the time and the
default sh on solaris works fine if I change this VT option.

It all depends on how your term program works.