Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site umcp-cs.UUCP
Path: utzoo!linus!decvax!harpo!gummo!whuxle!pyuxll!eisx!npoiv!npois!hogpc!drux3!ihnp4!zehntel!hplabs!hao!seismo!rlgvax!cvl!umcp-cs!israel
From: israel@umcp-cs.UUCP
Newsgroups: net.unix-wizards
Subject: the 'man' command
Message-ID: <3048@umcp-cs.UUCP>
Date: Sat, 8-Oct-83 01:10:58 EDT
Article-I.D.: umcp-cs.3048
Posted: Sat Oct  8 01:10:58 1983
Date-Received: Wed, 12-Oct-83 12:44:14 EDT
Organization: Univ. of Maryland, Computer Science Dept.
Lines: 25

I was checking the performance of an 'emacs' interface to the 'man'
command on our 4.1 bsd system, and I found something funny.

When 'man' is run normally, the first thing it does (after finding
the appropriate command and volume) is to see if it has already been
formatted in a /usr/man/cat<n> directory.  If so, it will use that
version instead of nroff'ing it, and pipe it thru 'more'.  If there
isn't one, or it's not up-to-date, it will make the formatted version.

However, I found by experimentation, (and confirmed it by examining
the code) that if the standard out of the man command is not a
teletype, but instead a file or pipe, not only does it skip piping
the output thru more (as one would expect it to) but it also skips
the check for the /usr/man/cat<n> file, and always nroff's the
source text.  In addition, it sends the output to its standard out,
but does not create or update the /usr/man/cat<n> file.

Does anyone know if there is some rational explanation for why it
does this, or has someone made the (probably fairly simple)
modifications for it to behave rationally?
-- 

~~~ Bruce ~~~
Computer Science Dept., University of Maryland
{rlgvax,seismo}!umcp-cs!israel (Usenet)    israel.umcp-cs@Udel-Relay (Arpanet)