[FOP-2860] [PATCH] BreakingAlgorithm causes high memory consumption

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 2.3
• Fix Version/s: 2.8
• Assignee: Simon Steiner
• Reporter: Raman Katsora
• Created: 16/Apr/19 07:06
• Updated: 14/Sep/22 08:53
• Resolved: 14/Sep/22 08:53

Description

when a single element (e.g. <fo:block>) contains a sufficiently large amount of text, the fo-to-pdf transformation causes very high memory consumption.

For instance, transforming a document with <fo:block> containing 1.5 million characters (~1.5Mb test-1500000.fo) requires about 3Gb of RAM.

The heapdump shows 27.5 million org.apache.fop.layoutmgr.BreakingAlgorithm.KnuthNode (~2.6Gb).

We start observing this issue, having about 300 thousand characters in a single element (test-300000.fo). But the high memory consumption isn't observed when processing 250 thousand characters (test-250000.fo).

Add <simple-line-breaking>true</simple-line-breaking> to fop.xconf for a lower memory method