[Tex/LaTex] Underfull \vbox – how to find offending page

vertical alignmentwarnings

I am compiling a large latex template, here, and always have a few underfull \vbox warnings.
I know that they're just issues with the spacing, discussion here, but is there a way to configure the warning message to tell me where the problem occurs?

That way I can easily find the problem pages, in both the PDF and the source, and decide if I need to fix it.

Best Answer

Most commonly under- and overfull \vboxes are associated with pages that don't have enough or too much content on them (as they are constructed as a vertical list of boxes). So, at the page-level, you can find the offending construction by looking at your log. Here's a quick example:

\documentclass{article}
\usepackage{lipsum}
\tracingpages1
\tracingoutput1
\begin{document}

\lipsum[1-50]
\noindent\rule{\textwidth}{1.1\textheight}
\lipsum[1-50]

\end{document}

The above code contains some dummy text followed by a single "paragraph" that is clearly bigger (vertically) than the text block (has a height of 10% more than \textheight). It also includes some tracing commands used for debugging. The .log shows

%% goal height=550.0, max depth=5.0
% t=10.0 g=550.0 b=10000 p=0 c=100000#
% t=22.0 g=550.0 b=10000 p=150 c=100000#
% t=34.0 g=550.0 b=10000 p=0 c=100000#
% t=46.0 plus 1.0 g=550.0 b=10000 p=250 c=100000#
% t=58.0 plus 1.0 g=550.0 b=10000 p=0 c=100000#
% t=70.0 plus 1.0 g=550.0 b=10000 p=0 c=100000#
% t=82.0 plus 1.0 g=550.0 b=10000 p=0 c=100000#
% t=94.0 plus 1.0 g=550.0 b=10000 p=0 c=100000#
% t=106.0 plus 1.0 g=550.0 b=10000 p=150 c=100000#
% t=118.0 plus 1.0 g=550.0 b=10000 p=0 c=100000#
% t=725.9478 plus 2.0 g=550.0 b=* p=150 c=*


Completed box being shipped out [10] []

%% goal height=550.0, max depth=5.0
% t=605.00336 g=550.0 b=* p=150 c=*#

Overfull \vbox (55.00336pt too high) has occurred while \output is active []



Completed box being shipped out [11] []

which we interpret as page [10] and being shipped out without problem. Then, after page [10] an Overfull \vbox occurred. Most likely this box is associated with page [11] as page [10] was already shipped out before this message occurred. Tracking the page construction of page [10] is not all that "user-friendly", as Knuth says himself. However, we use the following interpretation from the TeXBook (Chapter 15: How TeX Makes Lines into Pages, p 112):

The first line, which starts with %% is written when the first box or insertion enters the current page list; it shows the goal height and the max depth that will be used for that page (namely, the current values of \vsize and \maxdepth). [...] dimensions in the log file are always displayed in points.

The subsequent lines, which start with a single %, are written whenever a legal breakpoint is being moved from the list of recent contributions to the current page list. Every % line shows t, which is the total height so far if a page break were to occur, and g, which is the goal height; in this example g stays fixed [...], but g would have decreased if insertions such as footnotes had occurred on the page. The values of t are steadily increasing from 10 to 22 to 34, etc.; baselines are 12pt apart at the top of the page and 11pt apart at the bottom (where material is set in nine-point type).

We are essentially seeing one % line per hbox of text being placed on the current page. However, the % lines are generated by the penalty or glue items that follow the hboxes, not by the boxes themselves. Each % line shows also the badness b, the penalty p, and the cost c associated with a breakpoint; if this cost is the best so far, it is marked with a # sign, meaning that "this breakpoint will be used for the current page if nothing better comes along." [Wherever you have a badness of] b=10000, [...] they are so bad that TeX considers them indistinguishable; in such cases c=100000, so TeX simply accumulates material until the page is full enough to have b<10000. [...]. The notation b=* and c=* on the final line means that b and c are infinite; the total height [t] cannot be reduced to [g] by shrinking the available glue. Therefore the page is ejected at the best previous place [...].

It's therefore obvious that while reading content for page [10] TeX reached a point where there the total height could not be reduced to the goal height and the page was shipped out/ejected at the previous best break point, with that content spilling over to the subsequent page (page [11]). Page [11] only has a single, optimal (yet infinite) cost associated with the breakpoint, so the page is shipped out again... this time overfull as 605.00336pt is 55.00336pt too large for the 550pt page goal.

Related Question