Bug#659772: landscape sciposter + hyperref + latex + dvips = center-mirrored image
On Fri, Jul 13, 2012 at 12:39:27AM +0400, jaakov jaakov wrote:
> Let me notice that if you add setpagesize=false, the output is ok,
No, the output is truncated to A4 (or the default that is set in your
config.ps).
> but additionally providing -t landscape to dvips makes the output wrong
> again.
You need also "-t a0" to get A4 paper size.
> Probably, plain users should not care of such details or be
> guessing. By the way, why does hyperref have to set the page size by
> default?
Yes, already the LaTeX kernel should do this, but it was too early
frozen and the missing feature is left. In days of DVI the paper
size doesn't matter, because DVI does not know what this is.
But PS knows and especially PDF with its mandatory MediaBox and
four other optional boxes (CropBox, TrimBox, ArtBox, BleedBox).
Package hyperref is often used for PDF (links, bookmarks, ...),
therefore it has volunteered to do the job.
> If I get Heiko right, then, in high-level terms, another problem is the
> ambiguity in the specification of the dvi and postscript format (at least)
> with respect to the landscape or portrait format, as well as with respect
> to rotation/mirroring.
The problem with DVI is that it does not know a page size, it has
a funny origin with 1in margin in the left and at the top. But
it does not know a paper/media size or operations such as rotation.
This stuff is implemented via \special with driver dependent contents.
\specials for dvipdfmx might/will not work with dvips ...
The PostScript language allows the setting of the page size:
operator `setpagedevice', entry `PageSize', but entry `Orientation'
is only intended for roll-fed media. Also there is no easy way
to extract these data without a PostScript interpreter.
At least there are DSC-comments that allows the specification
of the page size and orientation. Thus there are several ways
for specifying (partly also version dependent) and some will
work on somewhere and others somewhere else.
Even worse if the width is greater than the height. That seems
to be the main problem here.
Yours sincerely
Heiko Oberdiek
Reply to: