Jeff,
 
Are you using Docbook 4 or 5? Also, are you using a current version of the  
XSL stylesheets? Is there a reference that tells you that Saxon doesn't 
resolve  XPointers?
 
Here is what I know from my own tools. Saxon/Xerxes and XSLTproc handle  
XPointers. I use them in every doc I generate. I use XSLTPROC for my 
production,  but maintain my Saxon tools as well because they will sometimes 
gives me 
more  info on a problem than xsltproc. I do use profiling and have not seen 
anything  like what you are describing lately.
 
When I was first setting up my tools, I had a dickens of a time using  
catalogs. Sometimes they would thrash for a while and error out. So I turned on 
 
catalog debugging and painstakingly went through each line the debugging  
produced and massaged it until I was sure of what it was doing. You may be  
seeing some of the same issues with your own catalog. If so, turn on the  
debugging and weed out the "circulars"
 
Now, if you are using Docbook 5, then I should bow out of this discussion  
since I have not made that leap yet.
 
Regards
Dean Nelson
 
 
In a message dated 3/12/2012 12:08:20 A.M. Pacific Daylight Time,  
[email protected] writes:

The problem I'm having  is unrelated to olinks, although I'm not sure what 
it  is.
 
All I know is that I  have problems generating larger (~400 pages) books 
with xsltproc using a  profiling stylesheet. If I generate a small book (50 
pages), it generates  without any problems, even if the book has XIncludes 
with XPointers,  profiling, and olinks.
 
If I generate a large  book, the catalog resolver chugs along endlessly to 
resolve generic  addresses to local paths. Not sure why. Perhaps there's 
infinite recursion  going on. xsltproc eventually runs out of memory. Curious 
to see if anyone  else observed infinite recursion problems with xsltproc. 
This doesn't happen  with Saxon-Xerces (but Xerces has other problems with 
XPointers).
 
After reading  this old post about infinite recursion 
(_http://mail.gnome.org/archives/xslt/2002-December/msg00026.html_ 
(http://mail.gnome.org/archives/xslt/2002-December/msg00026.html) )  I set the 
inherit.keywords parameter 
to 0, but that didn't solve the  problem:
 

<!-- Fixes a problem with xsltproc and  profiling -->
<xsl:param name="inherit.keywords">0</xsl:param>




Curious to hear how you got Saxon-Xerces to work with  XPointers, since 
Xerces doesn't support xml:id. Do you preprocess the book  with xmllint or 
another tool? 

Regards,
Jeff
 

 
____________________________________
 From: Bob Stayton [mailto:[email protected]]  
Sent: 2012-03-10 04:22
To: [email protected]; Jeff  Powanda; [email protected]
Cc:  [email protected]
Subject: Re: [docbook-apps] Re:  VirtualBox DocBook 5.0 VM



I'm curious too about the olink problem you mentioned.  Olink  collection 
and olink resolution should both take place after any XIncludes  have been 
processed and resolved, so I'm not clear how an XInclude with an  XPointer 
would create a problem.  Can you provide any more specifics  about that problem?
 
Bob Stayton
Sagehill Enterprises
[email protected]_ (mailto:[email protected]) 
 
 

----- Original Message ----- 
From:  [email protected]_ (mailto:[email protected])  
To: [email protected]_ (mailto:[email protected])  ; 
[email protected]_ (mailto:[email protected])  
Cc: [email protected]_ 
(mailto:[email protected])   
Sent: Saturday, March 10, 2012 3:23  PM
Subject: Re: [docbook-apps] Re:  VirtualBox DocBook 5.0 VM


Jeff,
Maybe I am not understanding the depth of the issue because I use  
XIncludes and XPointers and they work for me both in xsltproc and  Saxon/Xerces 
(I 
maintain both in my development environment).
 
I have also used profiling in the past and had no problem. However,  maybe 
you have come across an issue that I have not seen yet. 
 
Regards,
Dean Nelson
 
 
In a message dated 3/10/2012 1:06:20 P.M. Pacific Standard Time, 
[email protected]_ (mailto:[email protected])  writes:

I'm using  single-pass profiling. I've tried using two-pass profiling, and 
that  allows me to generate our larger books with xsltproc, but when I do 
that  the olinks to any XIncludes that use XPointers don't get resolved  
correctly. XPointers remain a problem for me whether I use xsltproc or  
Saxon-Xerces-J (which doesn't support xml:id syntax). Not sure what I  should 
do, 
other than avoid using XPointers.
 
Regards,
Jeff

 
____________________________________
 From: [email protected]  [mailto:[email protected]] 
Sent: 2012-03-10  07:52
To: Jeff Powanda; [email protected]
Cc:  [email protected]
Subject: Re: [docbook-apps]  Re: VirtualBox DocBook 5.0 VM



Jeff,
I use XIncludes, and not olinks or profiling (on the file I just  used). 
The only difference in the switches is that I use --novalid which  skips the  
skips the Dtd loading phase.
 
Are you using single pass profiling or two-pass?
 
Dean
 
 
In a message dated 3/9/2012 10:57:20 P.M. Pacific Standard Time,  
[email protected] writes:

Hmm, my docs are tiny  compared to that. Are you using XIncludes, olinks, 
and  profiling?

Here's how I'm calling xsltproc in my build.bat  file:

call xsltproc.exe 
--nonet 
--xinclude 
--stringparam profile.condition  %conditions% 
--stringparam olink.debug 0 
--stringparam collect.xref.targets "no" 
--output %output_file%  
--stringparam  insert.xref.page.number yes 
--stringparam  insert.olink.page.number maybe  
--stringparam  target.database.document %olink_file%  
--stringparam current.docid %doc_id% 
%xsl_file%  
%book_file%  

Regards,
Jeff

________________________________

From:  [email protected] [mailto:[email protected]] 
Sent: 2012-03-09  03:06
To: [email protected]; Jeff Powanda;  [email protected]
Cc:  [email protected]
Subject: Re: [docbook-apps] Re:  VirtualBox DocBook 5.0 VM


One more bit of info. I just ran a  1200 page doc (XP/2Gb/xsltproc) and it 
used about 230 MB of memory to  churn through that doc.

In a message dated 3/9/2012 2:44:36 P.M.  Pacific Standard Time, 
[email protected] writes:

Jeff,
I routinely process books over 400  pages (XP/2Gb/xsltproc) and rarely see 
a memory error. However, when I  DO see one, it is because I have done 
something wacky in the XSL or have  done something in the XML that was "legal" 
but not supported in the XSL.  Rare though.

What command line  switches are you using?

Regards,
Dean Nelson

In a message dated 3/9/2012 2:22:48 P.M. Pacific Standard Time,  
[email protected] writes:


I'm using the Windows version of  xsltproc.

xsltproc worked fine on our smaller manuals (~50 pages), but it wasn't  
able to handle our largest books (~400 pages). I need to be able to  build all 
the manuals on the writers' machines, which are Windows 7 or  Windows XP 
laptops with only 2 GB of RAM. Xsltproc couldn't seem to  handle those 
constraints. Saxon+Xerces can.

If other people are using  xsltproc under similar constraints on Windows 
and it's working fine,  perhaps I'm doing something wrong.

Regards,
Jeff



-----Original Message-----
From: Tom  Browder [mailto:[email protected]] 
Sent: 2012-03-09 09:31
To: Jeff  Powanda
Cc: Docbook Apps Help  list
Subject: Re: [docbook-apps] Re:  VirtualBox DocBook 5.0 VM

On Fri, Mar 9, 2012 at 11:25, Jeff Powanda  <[email protected]> wrote:
>  Ok, thanks. I switched from xsltproc a few years ago when it was  
> unable to process our largest  manuals (it ran out  of memory).

Was that a Windows version, or  *unix?

Size of manual?

I'm sure libxslt folks would love to hear about that whichever  system it 
was on.

-Tom

---------------------------------------------------------------------
To unsubscribe, e-mail:  [email protected]
For additional commands, e-mail:  [email protected]


=



Reply via email to