From: cwilson AT ee DOT gatech DOT edu (Charles Wilson)
Subject: Re: less, misbehaving
7 Oct 1998 14:53:14 -0700
Message-ID: <361A8AB2.1B8772E1.cygnus.gnu-win32@ece.gatech.edu>
References: <360FF514 DOT 34EEE46D AT ece DOT gatech DOT edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: gnu-win32 AT cygnus DOT com

Charles Wilson wrote:

> Well, I've tried many of the suggestions given concerning this problems, but
> the only thing that has changed is the description of the symptoms. Sorry
> for the length of this post; you can scan for '------' to get the major
> points.
> --------
> Recommendation 1 (Earnie Boyd): TERM=linux
>     I already had this.
> --------
> Recommendation 2 (Jeff Sturm & Larry Hall): get latest cygwinb19.dll from
> Sergey's site, and recompile ncurses and less myself, as both build easily
> OOB.
>     I already had latest dll, but ncurses didn't compile. I got 4.2 from the
> canonical site, ftp://www.clark.net/pub/dickey/ncurses/...  Here's a quick
> transcript snippet from configure / make:

<big snip>

> -------------
> Recommendation 3 (Jan Tomasek): type 'cat file.name | less'
>     this seems to work without bombing. But why? Other people don't have to
> resort to this, so what's wrong with my setup? (and Jan's, too)
> -------------
> Recommendation 4 (Earnie Boyd): use a different precompiled port of
> ncurses/less
>     haven't tried this yet, but will do so soon.

Tried it. didn't work.

<more snips>

Two things did work, however:
--------------
Recommendation 5 (Jonathan Storey): use old bash version
So, I replaced bash-2.0.2(1) with bash-2.0.1(2) Now less works just fine. The
only problem that could crop up (although I haven't seen it) - didn't
bash-2.0.1(2) have some wierd problems with random pauses? I seem to remember
seeing that on the mailing list archives, but it was before I started using
gnu-win (I've been using bash-2.0.2(1) exclusively from the beginning).

QUESTION: one of the main changes in these two version of bash was the switch
from fork() in the older version to spawn()/exec() in the newer one. (I think)
Could that be the difficulty here? I quote part of Jonathan's explanation:

"Whats happening here is that the process is requesting input from
the terminal i.e END: (less's case) and the shell thinks that the
process is in background. Under UNIX the message would be
Stopped: (tty input) or similar. Therefore I would conclude that
the new version of BASH.EXE is incorrectly flagging processes
as being in background albeit intermittently, although using
less(1) to output a file that is less than a screen full seems to
cause the problem to manifest itself consistently..."
----------------
The following over-engineered bashrc function works with bash-2.0.2(1)

# this deals adequately with the less misbehavior problems.
# downside: can only "less" one file at a time (I know, that means
# the prompts could do without the "file %i of %m" stuff, but I wanted
# to keep everything as close to the defaults as I could.)
# The +Gg options force less to run down to the end of the file, and back
# to the beginning..this allows less to know how many lines are in the
# file for displaying "line x of X" and percentage in the prompts.
# I don't find this distracting, since my K6-200 is fast enough that I
# don't see the display flickering to the end-of-file and back.
less() {
        local fn=$1
        shift
        cat $fn | /usr/local/bin/less -PM"$fn .?n?m(file %i of %m) \
...?ltline %lt?L/%L. :byte %bB?s/%s. .?e(END) ?x- Next\\: \
%x.:?pB%pB\\%..%t\$ -Pm?n$fn .?m(file %i of %m) ..?e(END) ?x- Next\\: \
%x.:?pB%pB\\%:byte %bB?s/%s...%t\$ -Ps?n$fn .?m(file %i of %m) ..?e(END) \
?x- Next\\: %x..%t\$" +Gg $*
}
export LESS="-isMd"

--Chuck


-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request AT cygnus DOT com" with one line of text: "help".