[XeTeX] A bug in (Xe)LaTeX?

[Bruno Voisin]

Le 18 août 05 à 21:55, Peter Dyballa a écrit :

>     \usepackage{geometry,graphicx}

I think you could remove the call of the geometry package here.

>     \addtolength{\latexitdepth}{2\fboxrule}    % correct correction  
> (was 0.72 pt before)

I'm a bit confused why \fboxrule is here (there's no framed box as  
far as I can tell).

>     \begin{document}
>     {\voffset=-1truein \hoffset=-1truein
>      \ifpdf
>        \pdfpageheight=\latexitheight
>        \pdfpagewidth=\latexitwidth
>      \else
>        \special{papersize=\latexitheight,\latexitwidth}%
>      \fi
>      \shipout\vbox{\scalebox{\latexitscalefactor}{\usebox 
> {\latexitbox}}}%
>     }%
>     \end{document}

For explaining the problem you report, I would imagine some sort of  
mixup between horizontal or vertical mode, yielding unwanted positive  
or negative vertical space somewhere. I've for a long time ceased to  
consider (La)TeX as deterministic regarding this issue (ie deciding  
whether it's in horizontal or vertical mode).

Does it change anything if you move the code setting the page  
dimensions (between {\voffset and \fi, and the final }%) before the  
\begin {document}? (I'm not sure you can do this with dvips, as I  
don't know whether the \special will be effective if it occurs before  
\begin{document}.)

And then, does it change anything if you add a \null (switching to  
horizontal mode) before \shipout, or before \scalebox, or if you  
replace \vbox by \hbox or \vcenter? I experienced bugs in LaTeX's  
page setting code, and had to add such a \null as the first element  
on a page, to position page elements precisely when designing a  
letterhead in LaTeX. If you don't do this, you'll never know whether  
or not the reference point at the beginning of a page is exactly  1in  
+ \topmargin + \headheight + \headsep + \topskip below the upper edge  
of the sheet of paper as it should.

That probably won't help, but who knows?

Bruno Voisin