Not sure that playing with the page-masters will really help. I don't think they have an influence here.
I've just committed a helper XSLT [1] which can split Luis' FO file at page-sequence boundaries. It reveals that one of the page-sequences alone makes up 18MB out of the 23MB of the full FO file. I guess that is what FOP chokes on: Just too much data in one page-sequence and FOP currently cannot free any objects while inside a page-sequence. [1] http://svn.apache.org/viewvc?rev=430134&view=rev I'd try to find way to further split up that large page-sequence. This should enable FOP to free memory and handle this file with less heap space. On 09.08.2006 20:13:43 Andreas L Delmelle wrote: > On Aug 9, 2006, at 19:27, Luis Ferro wrote: > > Hi, > > > It has LOADS of page sequences... it's a book with 2500 pages where > > from 3 or > > 4 pages, the "template" changes from one column page to 2 column > > page... > > > > Is there a better way of doing this "swap" of columns? > > Not that I'm aware of... (you're using conditional-page-masters, right?) > Can you do us a favor and try if using just one simple-page-master > makes a difference? Does it also consume a large amount of memory? > > Tinker a bit with the properties on each of the page-masters, see if > that changes anything... Never mind the output, it's simply to narrow > down the searching area. If the test succeeds with one simple-page- > master, or with different settings we'll get a better idea of where > to start looking. > > Thanks in advance! > > > > > Right now... with 477 pages, to render it ocupies 1.1Gb memory... (my > > machine has a top of 1.5Gb adressable to java...)... > > > > Will try now to simplefy everything as best as i can... > > Jeremias recently added a MemoryEater to the trunk with which we can > test, using one of your FO fragments and copying it a given number of > times. > Choose one representative page-sequence, and the structure of your > conditional-page-master-alternatives, post them in a Bugzilla --so > only those people that are interested need to download it-- and we'll > have a look. > > > How can i test if there is memory leaks somewhere (i'm a programmer > > but i'm > > very very green in java)? > > There is no easy way, I'm afraid. You can use a profiling tool one > one of the sessions --the JDK comes with some profiling facilities, > if you're a console-geek ;)-- to have a look at what the reference > trees in the heap look like at a certain point in the process, but > you'd still need some basic understanding of the process to figure > out which active references are totally unnecessary. > > If you're willing to invest time in this, of course you'd be welcome > to do so. If you have any questions or remarks, or need help > interpreting the results of a profiling session, just direct them to > fop-dev, or use Bugzilla to track the issue. > > Thanks again! > > > Cheers, > > Andreas Jeremias Maerki --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
