🚀 go-pugleaf

   From: George Young <young@vlsi.ll.mit.edu>
   Date: Tue, 29 Aug 89 13:15:49 EDT

   This is a bug report for bash version 1.02 .  Environment:
   TARGET:	SUN3	(sun 3/50, 12 Mbyte diskless)
   OS:	SUNOS3	(actually 3.5)
   BISON:	bison
   CC:	gcc -traditional -O -g

   Make ran without problems.  'bash' runs fine except:
	   When I type a ^D alone on a line, bash crashes with a segmentation
	   violation.  (If I exit bash with "exit" everthing's fine).
   gdb trace looks like:

   Program received signal 11, Segmentation fault
   0x2c390 in strlen ()
   (gdb) where
   #0  0x2c390 in strlen ()
   #1  0x451e in shell_getc () (parse.y line 393)
   #2  0x497e in read_token (command=0) (parse.y line 1027)
   #3  0x485a in yylex () (parse.y line 961)
   #4  0x34f6 in yyparse () (bison.simple line 214)
   #5  0x2974 in read_command () (shell.c line 634)
   #6  0x2852 in reader_loop () (shell.c line 574)
   #7  0x26a0 in main (argc=1, argv=(char **) 0xefffcf8, env=(char **) 0xefffd00) (shell.c line 465)

   This seems to be happening at line 434 of parse.y, when the
   'current_readline_line' arg to strlen has a bogus (-1 I think) value.
   The bug is quite repeatable.

   ..............................................................................
			   Unrelated bash comments:
   It would be helpful to have an 'autologout' timer,
   set by some variable.  I know this is only effective if one is sitting at
   a shell prompt, but lots of people leave a terminal sitting at a shell prompt
   for days (or weeks!).

   Documentation:  a full reference manual would be nice, when you have time, but
   in the mean time a heavily commented '.bashrc' file with lots of goodies would
   be very helpful, especially for users of '[t]csh' (like me) who would like to
   convert to bash.

   In general, this is a nifty shell; I have adopted it as my preferred shell on
   all our machines.
   George Young,  Rm. B-141		young@vlsi.ll.mit.edu
   MIT Lincoln Laboratory			young@ll-vlsi.arpa
   244 Wood St.				[129.55.11.7]
   Lexington, Massachusetts  02173		(617) 981-2756

   From: George Young <young@vlsi.ll.mit.edu>
   Date: Tue, 29 Aug 89 13:15:49 EDT

   It would be helpful to have an 'autologout' timer, set by some
   variable.

I have heard this request before, but I have not yet heard a good reason
for having it.

   I know this is only effective if one is sitting at a shell prompt,

Why?  It could just be based on idle time.

   but lots of people leave a terminal sitting at a shell prompt for
   days (or weeks!).

So what?  They are not taking any cycles by just sitting there.  They
would be taking up more cycles if the shell was getting alarm clock
signals telling it that `autologout' time had passed.

Brian

In article <8908291936.AA29762@aurel.caltech.edu> bfox@aurel.caltech.edu writes:
>   From: George Young <young@vlsi.ll.mit.edu>
>   It would be helpful to have an 'autologout' timer, set by some variable.
>   ...
>   I know this is only effective if one is sitting at a shell prompt,
>
>Why?  It could just be based on idle time.

I have yet to see an idle daemon that really works.  The ones I've tried (and
I have tried some) have always ended up kicking people off while they're in
the middle of editing.  Some programs read and write to /dev/tty and don't
update the modification time on the terminal.  People using the most popular
screen editor round here appear to be idle all the time they're in it, no
matter how fast you're typing.  The only record in the process structure of
how idle a process is has a very low limit (2 min ?) after which it is just
considered idle, so even chasing down the process tree is not good enough.

>   but lots of people leave a terminal sitting at a shell prompt for
>   days (or weeks!).
>
>So what?  They are not taking any cycles by just sitting there.

But they are leaving themselves open to attack.  I would like an autologout
feature, I do use tcsh's and I have modified csh here to timeout in the same
way (I've also modified the action of them both so that they give a warning
"press return to stay logged in" and redraw the input line 2/3rds of the way
through the timeout period).  This is not a means of restriction, but some
form of protection for the new or forgetful user.  New users get it set for
them in their default .cshrc.  They can disable it by unsetting a variable.
They can suspend it temporarily by typing `cat', and then ^D when they've
finished whatever they were doing.

Rob
-- 
UUCP:   ...!mcvax!ukc!warwick!cudcv	PHONE:  +44 203 523037
JANET:  cudcv@uk.ac.warwick             ARPA:   cudcv@warwick.ac.uk
Rob McMahon, Computing Services, Warwick University, Coventry CV4 7AL, England