Re: Troublesome checkwinsize (none) behaviour

[Linda Walsh]

werner@suse.de wrote:
> Description:
>         As subject says:
>           * If checkwinsize is _not_ set the bash does not update the
>             internal LINES and COLUMNS varaible during a command/job is
>             running in foreground on same terminal.
---
Right.

>           * If checkwinsize is set bash/libreadline does export LINES and
>             COLUMNS breaking ncurses/TIOCGWINSZ based commands/jobs.
Wrong.

Bash doesn't export them:
I have checkwin on:

  shopt|grep checkwin
  checkwinsize    on

But neither ROWS nor COLUMNS are set in the env:
  printenv|grep -P 'ROWS|COL'
  LS_COLORS=
  LC_COLLATE=C
  COLORTERM=1
  declare -i w=COLUMNS/2;

Note -- "/etc/rc.status on Opensuse *does* export COLUMNS for all processes.
It also sets up it's own signal handler which re-exports them on any
window change:

  # Seek for terminal size and, if needed, set default size
  rc_lc () {
      if test -n "$REDIRECT" ; then
          set -- $(stty size < "$REDIRECT"  2> /dev/null || echo 0 0)
      else
          set -- $(stty size 2> /dev/null || echo 0 0)
      fi
      LINES=$1
      COLUMNS=$2
      test $LINES   -eq 0 && LINES=24
      test $COLUMNS -eq 0 && COLUMNS=80
      export LINES COLUMNS
  }
  trap 'rc_lc' SIGWINCH
  rc_lc

----
I ran into the above when writing scripts because COLS/LINES would often be
undefined, so I had to protect many of my start scripts:

  B4ops="${-//[^eu]/}" ; set +e +u        # turn off strict fails
  . /etc/rc.status                        #safely include rc.status
  [[ $B4ops ]] && set -$B4ops

It would seem any process launched by init would have at least
COLUMNS in it's environment unless the ENV had been explicitly cleared --
and I know for interactive sessions, it is not.

As I found startup functions that dirtied the namespace with names like:

  path ()
  {
      command -p ${1+"$@"}
  }
---
Not even putting an underscore in front or back of it.  'path' is a
not an uncommon name for shell scripts to use.

Also, I assume you know that suse scripts export COLUMNS in places
like  /etc/profile, /etc/csh.login and /etc/ksh.kshrc...

Perhaps one of those is propagating to the error cases mentioned?