%% LblChng1.sty--tell about the first changed LaTeX label.
\def\fileversion{0.32c} \def\filedate{2005/10/01}
%% Copyright (C) 2005 Uwe L\"uck
%% Munich, http://contact-ednotes.sty.de.vu.
%% Current maintainer is Uwe L\"uck.
%% This program can be redistributed and/or modified under the terms
%% of the LaTeX Project Public License distributed from CTAN
%% archives in directory macros/latex/base/lppl.txt; either
%% version 1.3a of the License, or any later version.
%% We did our best to help you, but there is NO WARRANTY.
%% * SUMMARY: *
%%
% This package is intended to be a (primitive) tool for
% fixing oscillating page breaks -- a problem that especially
% sometimes occurs in typesetting critical editions. Thus it
% addresses especially users of LaTeX packages lineno, ednotes,
% and ledmac (maybe even manyfoot and numline). When it has been
% invoked and some label(s) get an entry in the .aux file
% differing, with respect to page number, to that of the
% previous run, the first such label is reported on screen (and
% in the .log), including the differing page numbers. This
% should help in finding the first unstable page -- which is
% vital. The file also contains rather detailed explanations and
% discussions.
%
%% * CRITICAL EDITIONS, PACKAGES: *
%%
% For informations about macro packages mentioned in the present
% text, or about critical edition typesetting in general, I
% refer you to the web page http://ednotes.sty.de.vu .
%
%% * THE PROBLEM TO BE SOLVED: *
%%
% Sometimes it happens that LaTeX's final warning "Label(s) may
% have changed. Rerun to get cross-references right." appears at
% *every* run, i.e., rerunning doesn't help. Or instead, it may
% be noticed that certain cross-references keep being wrong.
% The only reason I know of is that the runs oscillate between
% two decisions about what material should be put onto a certain
% page. And the only reason I know for this is that the space
% consumed by some cross-reference depends on the page breaking.
% I know this problem from critical edition typesetting.
%
%% * OUTLINE of the SOLUTION offered here: *
%%
% I simply propose that the user makes up her mind on the nasty
% page breaks and forces some at a certain lines, or contributes
% her opinion, through \pagebreak (or maybe even \nopagebreak, I
% have no idea whether this can be useful). -- A float (`figure'
% or `table') may contribute to the problem as well, so you may
% move its LaTeX code or change the placement options (through
% the brackets argument).
%
% Typically, one position of such oscillating causes further
% page breaking oscillations for many ensuing pages, yielding a
% bunch of wrong cross-references and countings. Of course, the
% user then should settle the first occurrence of oscillating
% before bothering with others -- all the rest of them may
% disappear when the first one is fixed.
%
% One problem with this solution ("?") is that it may be
% bothersome to find out where the first oscillation occurs
% indeed (by looking through the .dvi's or .pdf's). The present
% package tries to help with this.
%
% With Peter R. Wilson's ledmac, you have the "ballast"
% procedure for influencing page breaks. (It is due to John
% Lavagnino, Dominik Wujastyk, and Wayne Sullivan, inherited
% from EDMAC). It saves you from looking for good positions for
% \pagebreak's. I might offer something similar (with
% lineno.sty/ednotes.sty). However, I would like to object that
% the "ballast" procedure chooses between two page break
% possibilities rather "by chance" (reinforcing the first break
% idea encountered while the other idea might turn out to be
% better). But I must concede: One of the two breaks between the
% runs oscillate may turn out not to work, or even both of them
% -- when the cross-references have been adjusted to that
% decision. The "ballast" procedure may then save some time for
% you. On the other hand, if *both* decisions turn out not to
% work, I wonder whether the "ballast" method will certainly
% find a working one.
%
%% * USAGE: *
%%
% 1. \usepackage{lblchng1} in every LaTeX document that you
% create, or in every critical edition LaTeX document that you
% create, or only when you run into the problem of not settling
% label definitions. (Hope that you then remember that there is
% such a package!)
%
% 2a. You will be told about the page numbers where some \label
% (or, with Stephan I. B\"ottcher's lineno.sty or its extension
% ednotes.sty, some \linelabel) appears. Observe the page breaks
% in the .dvi's or .pdf's in the corresponding region. Note that
% the passage which causes the oscillation may be a number of
% pages earlier than the page number mentioned in the warning.
% (Sorry, this is the price of cheapness.) You can get rid of
% the warning without fixing the real problem -- fixing a late
% page break preceding the \label -- but this may cause badly
% filled pages earlier. So it will be wise to identify the very
% first passage causing oscillating page breaks. -- Another
% conjecturing hint: According to the diagnosis that I delivered
% above, the first oscillation happens near a cross-reference
% whose length varies due to the oscillation. So look for some
% \ref (or \lineref) referring forward(!). If you are lucky, the
% \ref refers to the label whose change has been warned about.
%
% 2b. If you are working with *ednotes.sty*: ednotes.sty v1.1
% does something similar with internal informations about each
% note, and it delivers a similar message on screen and in .log.
% It reports on three page numbers. However, the first
% oscillation happens (I think) at most one page before the
% reported ones.
%
% 3a. As indicated above, try to force a certain page break at
% this passage through \pagebreak (or \nopagebreak?) -- or
% change the position of a `figure' or `table' in code or in
% output (brackets argument), if it seems to contribute to the
% problem. Concerning \pagebreak (\nopagebreak), I personally
% also think of using the optional argument in the brackets,
% while I feel unable to judge the value of this. -- I also
% recommend that you use \typeout near the \pagebreak or so to
% remind yourself of the \pagebreak -- in case the \pagebreak
% after some changes suddenly produces a very bad page break
% and you don't examine the output with sufficient carefulness,
% being in a hurry, e.g.
%
% 3b. With *ledmac*, you may experiment with the "ballast"
% amount until that first oscillation stops.
%
%% * A minor WARNING: * (LaTeX version)
%%
% A remark as of 1995/07/13 in ltmiscen.dtx makes me wonder
% whether the present package works with older versions of
% LaTeX.
%
%% * CRITICAL EDITIONS: * (lineno.sty, ednotes.sty, ledmac)
%%
% 1. With Stephan I. B\"ottcher's lineno.sty and its extension
% ednotes.sty, oscillating page breaking also results in wrong
% pagewise numbering of lines and in wrongly placed-and-numbered
% footnotes. I am thinking of refining lineno.sty in addition to
% the present package. On the other hand, I conjecture that
% wrong pagewise numbering is *always* (at least sufficiently
% often) closely accompanied by oscillating cross-references, so
% the present package may offer sufficient help. -- Well, sorry,
% this may remain being true with respect to lineno.sty, but
% since March 2005, ednotes.sty treats its internal line number
% references in a way that makes them "invisible" to
% lblchng1.sty. This is very necessary for huge editions with,
% say, 10.000 notes or more, for a memory capacity reason.
% ednotes.sty uses an own procedure to detect page number
% changes, to which lblchng1.sty adds perhaps only little.
%
% 2. I guess that even ledmac users may find the present package
% useful, to find out where they have to watch when trying to
% adjust the "ballast" appropriately. However, its usefulness
% depends on how dense LaTeX \label's appear in the edited text.
% Indeed, it is the pages with the numerous critical footnotes
% fall where oscillations are likely to occur. \label's may be
% here when the comments refer to certain sections etc. of the
% edited text, OK, but will they? The line number references
% preceding the single notes (just on these "critical" -- in
% two senses -- pages) might be much more helpful, yet they are,
% with ledmac.sty, not implemented by \label. And it is wise not
% to implement them by \label, similarly as with ednotes.sty
% above. (ednotes.sty still uses \label, to be sure, but it
% redefines \newlabel in a way that is specific to ednotes.sty's
% way of storing data about the notes.)
%
%% * WONDERING REMARK: *
%%
% Oscillating label definitions do not necessarily result in
% anything wrong. The page number in the label definition may
% oscillate while not being used by any \pageref. -- I have no
% idea at present what this tells us to do if it happens so ...
%
%% * ACKNOWLEDGEMENTs: *
%%
% This package is due to Roy Flechner's confidence in our
% ednotes.sty and to frightening observations he reported to
% us. Thanks also to Peter Wilson for moral support.
%% Please send comments via http://contact-ednotes.sty.de.vu.
%% * IMPLEMENTATION: * %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\NeedsTeXFormat{LaTeX2e}[1994/10/30]
%% <- \@onelevel@sanitize; also \CheckCommand \reserved@a.
\ProvidesPackage{lblchng1}[\filedate\space\fileversion\space
^^Jexhibiting first changed label]
%% * Main idea: *
% Testing on label changes is done by \@testdef -- at least I
% assume this here (and I will somewhat check it). I refine it
% so it shows the label name and the page numbers -- clearly,
% what could I do else? (Well, I could redefine \label so ...)
%
%% If LaTeX's \@testdef changes, the present package may better
%% be updated:
\CheckCommand*\@testdef[3]{%
\def\reserved@a{#3}\expandafter \ifx \csname #1@#2\endcsname
\reserved@a \else \@tempswatrue \fi}
\def\@testdef#1#2#3{%% This is what I expect LaTeX to use.
\@ifundefined{#1@#2}\@tempswatrue{%% Like LaTeX:
\def\@tempa{#3}%
\expandafter
\ifx \csname #1@#2\endcsname \@tempa
\else \@tempswatrue
%% v0.3: moved disabling \@newl@bel from here to below.
\fi %% But now:
%% To be sure, I additionally assume that the page number is the
%% second thing, and more. I have had bad experience with
%% hyperref and thus do not assume that there are just two
%% things in the last argument of \newlabel.
\def\@tempb##1##2##3\@nil##4{%
\def##4{##2}\@onelevel@sanitize##4}%
%% Well, why not use some initially defined \@LC@extractpno
%% instead of this \@tempb? %% TODO
%% -- With modern large and fast TeX versions, one should not
%% waste a second with such considerations. -- However:
%% different effects concerning compatibility.
\expandafter \@tempb \@tempa \@nil\@tempa
\expandafter\expandafter\expandafter
\@tempb \csname#1@#2\endcsname \@nil\@tempb
\ifx\@tempa\@tempb \else
\typeout{^^JPackage `lblchng1':^^J%
*** Page number of label `#2' is changing: ***^^J%
\space\space\space\space
was \@tempb, will be \@tempa.^^JWatch whether
this stops and references get right.}%
%% Further testing wouldn't change anything:
\def\@newl@bel{\@gobblefour\relax}\fi}}
%% Warning: this could create an incompatibility with some other
%% package or with some LaTeX version.
%% Again: this extra line doesn't pay well nowadays.
\endinput
VERSION HISTORY:
v0.1 2005/02/15 First release for Roy Flechner.
v0.2b 2005/02/16 Attempt at \getpagewiselinenumbers -- in
vain. Improved explanations.
v0.2 2005/02/17 Changed implementation back; but keep
completely disabling \newlabel after the
report. Only page numbers are reported.
(CTAN and ednotes.news received slightly
different explanations than Roy Flechner.)
v0.3 2005/02/17 \@newl@ble turned off with page no. change
only. Doc. warns on compatibility.
v0.31 2005/02/27 Additional line on screen.
v0.32 2005/03/08 Adapted explanations to ednotes.sty v1.1;
warned about usefulness with critical
editions. Changed some headings.
Acknowledgements.
v0.32a 2005/03/21 Minor corrections.
v0.32c 2005/10/01 Another typo fix.