[PDFBOX-5158] Infinite loop on corrupted PDF in 3.0.0-SNAPSHOT

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 3.0.0 PDFBox
• Fix Version/s: 3.0.0 PDFBox
• Assignee: Tilman Hausherr
• Reporter: Tim Allison
• Created: 08/Apr/21 16:54
• Updated: 18/Aug/23 05:46
• Resolved: 09/Apr/21 06:34

Description

I found a bunch of files that had a "read too many EOFs", which is a safety check we now do in TikaInputStream to identify parsers that read an EOF > 1000 times, which may be a sign of an infinite loop.

When I turn off this safety check in TikaInputStream, I get an infinite loop.

This is one of the triggering files: https://corpora.tika.apache.org/base/docs/commoncrawl3/OE/OELHPKYAQPDNDWC535NE23Z6FKYRMN7W

It's a truncated file from Common Crawl.

The stacktrace when this is thrown is:

afterRead:809, TikaInputStream (org.apache.tika.io)
read:82, ProxyInputStream (org.apache.commons.io.input)
<init>:113, RandomAccessReadBuffer (org.apache.pdfbox.io)
loadPDF:454, Loader (org.apache.pdfbox)
loadPDF:430, Loader (org.apache.pdfbox)
getPDDocument:189, PDFParser (org.apache.tika.parser.pdf)
parse:148, PDFParser (org.apache.tika.parser.pdf)
parse:288, CompositeParser (org.apache.tika.parser)
parse:288, CompositeParser (org.apache.tika.parser)
parse:150, AutoDetectParser (org.apache.tika.parser)
parse:157, RecursiveParserWrapper (org.apache.tika.parser)
getRecursiveMetadata:379, TikaTest (org.apache.tika)
getRecursiveMetadata:369, TikaTest (org.apache.tika)
getRecursiveMetadata:357, TikaTest (org.apache.tika)
getRecursiveMetadata:351, TikaTest (org.apache.tika)

Issue Links

• duplicates: PDFBOX-5543 pdfbox 3.0.0-RC1 | Loader.loadPDF(inputStream) is getting stuck for few pdf files