References: <5.1.0.14.0.20050417104145.0247e860@pop3.web.de>
User-Agent: Gnus/5.11 (Gnus v5.11) Emacs/22.0.50 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit
Message-ID:  <85y8bhem3i.fsf@gnu.org>
Date:         Sun, 17 Apr 2005 12:25:37 +0200
Reply-To: Mailing list for the LaTeX3 project
              <LATEX-L@listserv.uni-heidelberg.de>
Sender: Mailing list for the LaTeX3 project
              <LATEX-L@listserv.uni-heidelberg.de>
From: David Kastrup <dak@GNU.ORG>
Subject: Re: \output trick; Re: lineno.sty question
To: LATEX-L@listserv.uni-heidelberg.de
In-Reply-To:  <5.1.0.14.0.20050417104145.0247e860@pop3.web.de> (Uwe
              =?utf-8?Q?L=C3=BCck's?= message of "Sun, 17 Apr 2005 11:14:47
              +0200")
Precedence: list
Status: R

Uwe Lück <uwe.lueck@WEB.DE> writes:

> Dear Erik Luijten,
>
> At 01:00 16.04.05, Erik Luijten wrote:
>>2. As a physicist, I frequently use the widely-used package "revtex". I
>>found that this breaks lineno completely (both version 3 and version 4). I
>>was wondering whether there is a chance that lineno.sty can be made
>>compatible with revtex?
>
> This is due to a mistake in revtex's way of changing the output routine.
> After a \newtoks\output, the new output routine invokes \the\output,
> it should use a private string instead (\let\rev@output\output ...).
> lineno.sty tries the same trick, but due to revtex's \the\output,
> lineno's output routine is entirely ignored.
>
> This was pointed out by David Kastrup recently. I had made the
> mistake myself, and so has lineno.sty done up to now. David Kastrup
> called it a "standard mistake", I wondered who made it else, now
> I find it funny to see another instance.

Perhaps the most glaring instance: latex.ltx

\let\frozen@everymath\everymath
\let\frozen@everydisplay\everydisplay
\newtoks\everymath
\newtoks\everydisplay
\frozen@everymath = {\check@mathfonts
                     \the\everymath}
\frozen@everydisplay = {\check@mathfonts
                        \the\everydisplay}

I have reported this as a bug several years ago, IIRC including
examples of stuff that breaks because of this.

I think it was more or less declined to fix this because of "bug
compatibility".

If you take a look at amsmath.sty, it sometimes uses \everymath and
\everydisplay, sometimes \frozen@everymath and \frozen@everydisplay.
The resulting code is rather opaque, and it is hard to imagine that it
can be both correct and _recognizable_ as correct.

All of this would be unnecessary if latex.ltx redirected the hooks in
the proper manner described.

> "Mistake" may be inappropriate, but David Kastrup's proposal is
> a helpful convention.

A programming practise that will break if anybody else will use it
too, is a mistake.  Call it a violation of a programmer's
"Kategorischer Imperativ".

You can see that as a result amsmath.sty broke in several places, and
a fragile pseudofix has been applied at those places that actually
triggered a bug report up to now.

I think "mistake" is quite appropriate to characterize such coding
practices.

--
David Kastrup, Kriemhildstr. 15, 44793 Bochum