[poppler] Printing of certain PDF files does not work with "fit-to-page" because of wrong BoundingBox values in the PostScript

[Adrian Johnson]

On 28/10/15 04:03, William Bader wrote:
> 
> 
>> To: stefan.brandner at gmx.at; poppler at lists.freedesktop.org
>> From: ajohnson at redneon.com
>> Date: Tue, 27 Oct 2015 23:08:54 +1030
>> Subject: Re: [poppler] Printing of certain PDF files does not work
> with "fit-to-page" because of wrong BoundingBox values in the PostScript
>>
>> On 25/10/15 21:08, Stefan Brandner wrote:
>> > I want to summarize how the problem appears when printing the file
>> > Dirichlet.pdf with okular.
>> > Find the file at: https://bugzilla.opensuse.org/attachment.cgi?id=616147
>> >
>> > 1. with OpenSuse 13.1 and okular 4.11.2 the CUPS command was without the
>> > fit-to-page option
>> > following commands are executed:
>> > pdftops Dirichlet.pdf
>> > lpr Dirichlet.ps
>> > The file was printed unscaled but centered and the printout was useful.
>> >
>> > 2. with the actual OpenSuse 13.2 and okular 4.14.9 the CUPS command is
>> > now with fit-to-page and then the file is shifted to the right with half
>> > of the text cut
>> > this can be reproduced with following commands:
>> > pdftops Dirichlet.pdf
>> > lpr -o fit-to-page Dirichlet.ps
>> > This does not work since the BoundingBox values are incorrect.
>> > See explanation at CUPS documentation:
>> > http://www.cups.org/documentation.php/options.html
>>
>> I'm not sure which application is enabling the fit-to-page or why it
>> changed. The fit to page would be better performed in pdftops which
>> already has this capability. The problem with performing a fit-to-page
>> based on the PS output is the Adobe Document Structuring Conventions
>> define the BoundingBox to be the smallest box that encloses all marks on
>> the page.
>>
>> A PS file with correctly calculated PageBoundingBox comments is likely
>> have a different bounding box on each page depending on the content. If
>> fit-to-page is applied based on the bounding box, each page will have a
>> different scale factor applied. A blank page with just a page number at
>> the bottom will be scaled up to fit the page number to the entire page.
>> This is probably not what you want.
>>
>> The fit-to-page feature relies on the fact that a lot of applications
>> just set the bounding box to the page size since computing the ink
>> extents of all marks on the page is not trivial.
>>
>> If a future version of poppler tightened up the PageBoundingBox to make
>> it more correct you may find this workflow which relies on an incorrect
>> interpretation of the BoundingBox comment may no longer work for you.
> 
> 
> The problem is that the pdftops -paper option creates a
> PageBoundingBoxes that can point to empty space instead of to the image.
> It is not that the PageBoundingBox is too large but that it is
> completely wrong.

Yes, I understand the current bounding box is bogus and the patch fixes
it to ensure the bounding box does not truncate any of the page content.
That alone is a good enough reason to apply the patch.

But my point is the new bounding box still does not comply with the DSC
spec. A future version of poppler may make the bounding box "more
correct" and as a result the CUPS fit-to-page feature will not work the
same as it does with the current patch.


>  pdftops always writes the PageBoundingBox as 0,0 to
> the width,height of the image, but if pdftops is run with -paper, it
> shifts the image up and to the right to center it on the page but it
> does not make a corresponding shift to the PageBoundingBox.
> 
> 
> The issue is not setting the PageBoundingBox to cover the smallest
> rectangle necessary to cover the marks on the page.  An application that
> creates a document where a page has only a page number or a "this page
> intentionally left blank" notice will usually still write a normal
> page-sized crop box.  But if you have a document that was created for a
> small paper size like a note card and then you want to print it on A4,
> if you use pdftops -paper A4, pdftops will correctly center the note
> card pages on the A4 paper (and enlarge them if you use -expand), but it
> will leave the PageBoundingBox comments as a note card sized rectangle
> at the lower left corner, so viewers or applications that read the
> PageBoundingBox will look at empty space.
> 
> 
> Regards, William
> 
> 
>