bug#12646: 24.2.50; Eshell problem with mailx(1)

bug#12646: 24.2.50; Eshell problem with mailx(1)

[Stephen Berman]

Since updating my openSUSE system from 12.1 to 12.2, I cannot read email
by invoking mailx from the eshell prompt:

Welcome to the Emacs shell

/data/steve/lib/emacs/gtk $ mail
Heirloom mailx version 12.5 7/5/10.  Type ? for help.
"/var/spool/mail/steve": 1 message 1 unread
>U  1 jan.h.d@swipnet.se Sat Oct 13 23:53  109/4392  Re: bug#12612: 24.2.50; F
? 
WARNING: terminal is not fully functional
-  (press RETURN)
?  
At EOF
? 

This happens both with my Emacs build from the current trunk and with
openSUSE's official build of 23.3.

With openSUSE 12.1, which contains the same version of mailx as 12.2, I
also get the warning but after typing RETURN, the top of the mail is
displayed and I can scroll through it by repeatedly typing RET.  This is
openSUSE's official build of 23.3 and with a build of 24.1 (I can't test
the current trunk on openSUSE 12.1 without rebuilding Emacs, which is
currently rather inconvenient).  Both versions of openSUSE use the same
version of mailx, and in both the value of $TERM in eshell is dumb.  If
this is not an Emacs or Eshell problem, what could account for it?


In GNU Emacs 24.2.50.3 (x86_64-suse-linux-gnu, GTK+ Version 3.4.4)
 of 2012-10-14 on rosalinde
Bzr revision: 110543 handa@gnu.org-20121014080652-4o4ptkifsi2xqtk4
Windowing system distributor `The X.Org Foundation', version 11.0.11203000
System Description:	openSUSE 12.2 (x86_64)

Configured using:
 `configure '--without-toolkit-scroll-bars' 'CFLAGS=-g -O0''

Important settings:
  value of $LANG: en_US.UTF-8
  value of $XMODIFIERS: @im=local
  locale-coding-system: utf-8-unix
  default enable-multibyte-characters: t

bug#12646: 24.2.50; Eshell problem with mailx(1)

[Andreas Schwab]

Stephen Berman <stephen.berman@gmx.net> writes:

> Welcome to the Emacs shell
>
> /data/steve/lib/emacs/gtk $ mail
> Heirloom mailx version 12.5 7/5/10.  Type ? for help.
> "/var/spool/mail/steve": 1 message 1 unread
>>U  1 jan.h.d@swipnet.se Sat Oct 13 23:53  109/4392  Re: bug#12612: 24.2.50; F
> ? 
> WARNING: terminal is not fully functional
> -  (press RETURN)

This is a bug in less, which crashes here.  As a workaround, disable the
pager.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#12646: 24.2.50; Eshell problem with mailx(1)

[Stephen Berman]

On Sun, 14 Oct 2012 15:00:34 +0200 Andreas Schwab <schwab@linux-m68k.org> wrote:

> Stephen Berman <stephen.berman@gmx.net> writes:
>
>> Welcome to the Emacs shell
>>
>> /data/steve/lib/emacs/gtk $ mail
>> Heirloom mailx version 12.5 7/5/10.  Type ? for help.
>> "/var/spool/mail/steve": 1 message 1 unread
>>>U  1 jan.h.d@swipnet.se Sat Oct 13 23:53  109/4392  Re: bug#12612: 24.2.50; F
>> ? 
>> WARNING: terminal is not fully functional
>> -  (press RETURN)
>
> This is a bug in less, which crashes here.  As a workaround, disable the
> pager.
>
> Andreas.

Thanks for your explanation and your suggestion, but I have problems
with each and, even though this apparently is not an Emacs issue, I hope
you will indulge me a bit further:

First, regarding your suggested workaround, I'm not sure exactly what
you mean.  I tried `export PAGER=""' both from eshell and also from the
shell before starting Emacs, and neither changed the mailx problem in
eshell.  Then I tried `export PAGER="more"' and with that I could now
read mail with mailx from eshell; is that what you meant?  However, this
is an unsatisfactory workaround, since `more' is much worse than `less';
also, if this is only a problem with mailx, I wouldn't want to have
`less' replaced by `more' generally in eshell.  Or if you meant
something else, please say precisely what I have to do to disable the
pager.

Regarding your explanation, both my openSUSE 12.1 and 12.2 installations
contain the same version of less, less 444, yet the eshell problem with
mailx only happens in 12.2, so the bug must be elsewhere; is it in the
12.2 packaging (less-444-9.1.2.x86_64.rpm vs. less-444-3.2.1.x86_64.rpm
in 12.1)?  But if so, why is there no problem using less itself from
eshell in openSUSE 12.2?  And why does this bug only show up with mailx
in eshell (AFAIK)?  There is no problem using mailx with `M-x shell' or
`M-x term', nor from outside of Emacs (all with `less' as pager).  If
there is a bug in `less' in openSUSE 12.2, I should report it, but aside
from just repeating my Emacs bug report, I can't see what else to say;
if you can elaborate the issue, that would help.

Steve Berman

bug#12646: 24.2.50; Eshell problem with mailx(1)

[Andreas Schwab]

The bug is in less, but triggered by changes in ncurses.  You should be
able to get the same crash in 12.1 by exporting LINES=0 in the
environment.

To disable the pager in mailx set crt to a big value.  RTFM.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#12646: 24.2.50; Eshell problem with mailx(1)

[Stephen Berman]

On Mon, 15 Oct 2012 01:25:36 +0200 Andreas Schwab <schwab@linux-m68k.org> wrote:

> The bug is in less, but triggered by changes in ncurses.  You should be
> able to get the same crash in 12.1 by exporting LINES=0 in the
> environment.

Unfortunately I couldn't test this: I exported LINES=0 in the shell and
then started Emacs, but in Eshell `echo $LINES' returned 42; same when I
exported LINES=0 within Eshell and also when I used setenv, both before
and after starting Eshell.  I searched the Eshell manual but found
nothing about setting environment variables.

I'm still interested to know why I encounter the bug only in Eshell and
only with mailx, not with less itself; do you have any idea?

> To disable the pager in mailx set crt to a big value.  RTFM.
>
> Andreas.

Thanks.

Steve Berman

bug#12646: 24.2.50; Eshell problem with mailx(1)

[Andreas Schwab]

Stephen Berman <stephen.berman@gmx.net> writes:

> I'm still interested to know why I encounter the bug only in Eshell and
> only with mailx, not with less itself; do you have any idea?

$ stty rows 0

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#12646: 24.2.50; Eshell problem with mailx(1)

[Stephen Berman]

On Mon, 15 Oct 2012 12:39:53 +0200 Andreas Schwab <schwab@linux-m68k.org> wrote:

> Stephen Berman <stephen.berman@gmx.net> writes:
>
>> I'm still interested to know why I encounter the bug only in Eshell and
>> only with mailx, not with less itself; do you have any idea?
>
> $ stty rows 0

Do you mean that, as far as Eshell is concerned, the terminal always has
0 rows?  At least `stty -a' in Eshell always shows "rows 0", no matter
what value I give to N in `stty rows N' at the eshell prompt, whereas in
a linux console, e.g. KDE Konsole, `stty rows 0' does indeed change the
rows terminal characteristic and prevents reading mail with mailx, just
as in Eshell.  But e.g. `stty rows 37' in Konsole -- but not in Eshell
-- restores the number of rows and the use of mailx (however, `stty rows
0' in Konsole also appears to suppress echoing of keys, and e.g. `stty
rows 37' does not restore that).  If my understanding of your answer is
correct, then it seems to me that this indicates an Eshell doc bug and
should at least be noted in PROBLEMS.  Of course, the Eshell manual has
always been mostly empty, so it may be regarded as a known bug, but this
particular case (the change in behavior in mailx when invoked from
eshell) is not obvious and likely to catch others who, like me, are not
well-acquainted with how such properties of ttys interact with commands
like mailx.

Steve Berman

bug#12646: 24.2.50; Eshell problem with mailx(1)

[Andreas Schwab]

Eshell creates a new tty for each command, so any setting will never be
persistent.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#12646: 24.2.50; Eshell problem with mailx(1)

[Stephen Berman]

On Mon, 15 Oct 2012 16:06:14 +0200 Andreas Schwab <schwab@linux-m68k.org> wrote:

> Eshell creates a new tty for each command, so any setting will never be
> persistent.

I tried tracing the eshell call chain after invoking mailx, but I
couldn't find where the tty gets created; can you point me to the code?
Also, why is the tty always created with the rows terminal
characteristic set to 0?  Does Eshell do this (if so, where?) or is it a
system or kernel default that Eshell doesn't change?  Thanks.

Steve Berman