Re: [cjk] Prepend \leavevmode to address@hidden, address@hidden, address@hidden, etc

[Werner LEMBERG]

From: Qing Lee <address@hidden>
Subject: Prepend \leavevmode to address@hidden, address@hidden, address@hidden, 
etc
Date: Mon, 02 Jun 2014 08:03:14 +0800

> The motivation for this request is based on the bug of the example
> in attached file.

Thanks for the report.

> I am sorry that this example is quite strange.  But this bug is not
> very easy to reproduce.  You can open `test-cjkutf8-cmap-bug.pdf'
> via a PDF browser and copy some text from it.  Then paste them to
> some suitable place, you will find that the result is very weird.
> For example, the result of `一一 XYZ 2000` is `一一 乘乙乚 串丰丰丰
> `.  The result of CJK characters is right, but others is wrong.

I've simplified your example to

  \documentclass{article}

  \usepackage{CJKutf8}


  \begin{document}
  \begin{CJK}{UTF8}{gbsn}

  \vspace*{19cm}

  a

  一

  \end{CJK}
  \end{document}

If you now cut and paste `a', the (incorrect) result is `乡'.
However, if you replace value `19' with, say, `18.5', everything is
just fine.

> It seems that \ToUnicode cmaps is also add to none-CJK font (here is
> \OT1/cmr/m/n/10).  This is obviously wrong.  I find the reason is
> that address@hidden is also defined in `\output` in this example.
> This led address@hidden to the wrong branch in `\output`.

Indeed, it's a bad interaction with the output routine.  With value
`19', an additional blank page gets inserted at the very beginning of
the above sample for no obvious reason.  This is a hint that \output
is seriously confused.

Under normal circumstances, the bug doesn't show up, as far as I can
see.  With `normal' I mean that LaTeX's output routine has a chance to
correctly finish the current page, without being fooled by strange
vertical offsets that makes text appear below the text's bottom
margin.

I guess it should be straightforward to fix your document to avoid the
problem by slightly modifying vertical distances, or adjusting margins
or paper dimensions.

Unfortunately, I don't have time currently to debug this in great
detail, but using the above minimum example you might ask for help on
StackExchange, for example.

> If I add `\leavevmode` in the very beginning of the definition of
> address@hidden, say
>
>     address@hidden@altchar##1##2##3{
>       \ifvmode\leavevmode\fi  %% added by me
>       {\ifnum ##2 < address@hidden
>          address@hidden
>        \else
>        ......
>
> The bug is gone and everything seems be OK.

Before adding such a fix it would be nice to really understand what's
going on...


    Werner