longtabu and floats: wrong table breaks on pages with floats

Question

I noticed that the tabu package seems to have problems with its longtabu environment, if it is placed on a page, where a float already is placed.

Consider the following MWE:

\documentclass[10pt]{article}
\usepackage{longtable,tabu}
\usepackage{lipsum}

\begin{document}

    \begin{table}
        \begin{tabular}{ccc}
            Row 1 & Row 1 & Row 1 \\
            Row 2 & Row 2 & Row 2 \\
            Row 3 & Row 3 & Row 3 \\
            Row 4 & Row 4 & Row 4 \\
            Row 5 & Row 5 & Row 5 \\
            Row 6 & Row 6 & Row 6 \\
            Row 7 & Row 7 & Row 7 \\
            Row 8 & Row 8 & Row 8 \\
            Row 9 & Row 9 & Row 9 \\
            Row 10 & Row 10 & Row 10 \\
        \end{tabular}
    \end{table}

    \lipsum[1-2]

    \begin{longtabu}{X[l]X[c]}
        Test & Test
    \end{longtabu}

    \lipsum[10]

    {
        \footnotesize
        \begin{longtabu}{|X[-4,l]|X[r]|X[r]|X[r]|}
            \hline
            Col 1 & Col 2 & Col 3 & Col 4 \\\hline
            Lorem ipsum dolor sit & 0.000019 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000012 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000173 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000011 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.012992 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000031 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000068 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000018 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.029689 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000055 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000004 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 28.276892 & 0.405603 & 1.560010 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.957158 & 0.234002 & 0.702005 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000011 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000002 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000013 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000336 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000004 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000003 & 0.000000 & 0.000000 \tabularnewline\hline
            Lorem ipsum dolor sit & 0.000003 & 0.000000 & 0.000000 \tabularnewline\hline
        \end{longtabu}
    }

\end{document}

Note that the last table is broken in a wrong way at the bottom of page 1. I guess that the wrong offset for the table break is exactly the vertical dimension of the first table, which is enclosed in a floating environment.

However, I don't know how to fix the wrong table break. I have many pages containing floating tables, but at some places I would like to have long tables with longtabu, as exactly as in the MWE.

Can anybody tell me how to fix the wrong table break?

Answer 1

There have been some bugs in that area, although I thought they were fixed. If its reproducible just using longtable rather than the tabu version, please

latex latexbug.tex

and follow the prompts to submit a bug (category tools) although no promises getting this right is hard and changing the code after all this time is hard even if it is "wrong". Putting \clearpage before the longtable will flush the float and ensure things are in sync, although can cause a premature page break of course, but if you need the document in a hurry that may be the best course of action.

Meanwhile I or someone else here may spot a fix directly in your example....

UPDATE: FIX

If you apply the changes suggested here

http://www.latex-project.org/cgi-bin/ltxbugs2html?pr=tools/3512

then your example works as expected.

I seem to recall there was a reason that change wasn't just immediately applied at the time, but I don't see anything in the log, I'd need to investigate, but hopefully this will get you working again.