PDF layout errors in headless mode

Hello all, 

I am running a document conversion server running
OpenOffice.org in headless 
mode (using a virtual framebuffer) and a little Python
script[1] to convert 
documents from ODT to PDF. However, I keep running into
layour errors when 
converting documents. When I "manually" convert
these documents to PDF they 
all work fine however. 

[1] 
http://www.artofsolving.com/opensource/pyodconverter 

I have uploaded some files to my webserver: 

The original ODT:
http
://www.jejik.com/sander/temp/foobar.orig.odt 

A PDF generated manually (looks good):
ht
tp://www.jejik.com/sander/temp/foobar.normal.pdf 

A PDF generated headless (looks bad. See page 2):

http://www.jejik.com/sander/temp/foobar.headless.pdf 

Just to make sure that it wasn't any configuration setting,
I ran the 
conversion server on the same computer as my regular
OpenOffice.org, using 
the exact same configutaion directory (~/.openoffice.org2 on
my Linux box). 

Anyone know what's going on here and how I can fix this
issue? Kind regards, 

-- 
Sander Marechal

------------------------------------------------------------
---------
To unsubscribe, e-mail: dev-unsubscribeapi.openoffice.org
For additional commands, e-mail: dev-helpapi.openoffice.org

Hi Sander,

(I'm the author of that Python script.)

Sander Marechal wrote:
> I am running a document conversion server running
OpenOffice.org in
> headless mode (using a virtual framebuffer) and a
little Python
> script[1] to convert documents from ODT to PDF.
However, I keep
> running into layour errors when converting documents.
When I
> "manually" convert these documents to PDF
they all work fine however.
The layout problem I can see in your "headless"
PDF is that on the
second page there is a table but (party) hidden under the
line and pie
charts, while in your "normal" PDF the table is
entirely visible above
the charts.

I tried converting your ODT but I always get the same output
both with
the script and manually. However, with OOo 2.2 shipped by
Ubuntu, the
layout is wrong in both cases, in fact shows incorrectly in
Writer as
well; while on the same machine with OOo 2.2.1 from
openoffice.org the
layout is always correct.

Not sure why that should be the case, but maybe this can
help in finding
out.

Kind regards

Mirko

------------------------------------------------------------
---------
To unsubscribe, e-mail: dev-unsubscribeapi.openoffice.org
For additional commands, e-mail: dev-helpapi.openoffice.org

Mirko Nasato wrote:
> (I'm the author of that Python script.)

I know, we talked before a few months ago

> The layout problem I can see in your
"headless" PDF is that on the
> second page there is a table but (party) hidden under
the line and pie
> charts, while in your "normal" PDF the table
is entirely visible above
> the charts.

The odd thing is that the line and pie chart are themselves
in a
separate one row, two column table that comes after the
table they are
obscuring. The two charts are anchored to the paragraphs
inside the two
cells. If you open up the content,xml you can see it. I see
no reason
why they should float to the top.

> I tried converting your ODT but I always get the same
output both with
> the script and manually. However, with OOo 2.2 shipped
by Ubuntu, the
> layout is wrong in both cases, in fact shows
incorrectly in Writer as
> well; while on the same machine with OOo 2.2.1 from
openoffice.org the
> layout is always correct.

I am running openoffice 2.0.4 from Debian Etch (stable). I
tried
updating to 2.2 from backports.org but I got the same result
that you
described: wrong on both cases. There were a few more errors
in 2.2 (bad
handling of the DIN fonts, causing spaces to appear where
there should
be none) so I reverted back to the stable 2.0.4 shipped with
Debian.

To make the situation even more complicated is that a very
similar
document does work. I have a single ODT with the same
contents that
works fine with OOo 2.0.4 headless. I wanted to try
splitting up the
sections and using an ODM, then export the ODM to ODT and
break the
links*. The result is the ODT I posted that goes wrong.

All in all, a messy situation. I'll see if I can find Debian
packages
for 2.2.1 and try that.

-- 
Sander Marechal

* These documents are generated automatically by a system
not unlike
your JODReports, but based on XSLT. I wanted to be able to
generate the
parts separately, but still end up with one ODT in the end.


------------------------------------------------------------
---------
To unsubscribe, e-mail: dev-unsubscribeapi.openoffice.org
For additional commands, e-mail: dev-helpapi.openoffice.org

Hi Sander,

Sander Marechal wrote:
> Mirko Nasato wrote:
>   
>> (I'm the author of that Python script.)
>>     
>
> I know, we talked before a few months ago
>
>   
Yes I remember 

> All in all, a messy situation. I'll see if I can find
Debian packages
> for 2.2.1 and try that.
>   
You can always download the RPMs from openoffice.org and
convert them to
DEBs with alien.

In fact, I simply uncompress the RPMs with cpio and install
them to a
local directory, not managed by the package manager. This
way I can keep
different versions of OOo installed for testing.

Kind regards

Mirko

------------------------------------------------------------
---------
To unsubscribe, e-mail: dev-unsubscribeapi.openoffice.org
For additional commands, e-mail: dev-helpapi.openoffice.org