Tilman Hausherr created PDFBOX-2762:
---------------------------------------
Summary: remove parseCOSStream() call from PDFStreamParser
Key: PDFBOX-2762
URL: https://issues.apache.org/jira/browse/PDFBOX-2762
Project: PDFBox
Issue Type: Task
Components: Parsing
Affects Versions: 2.0.0
Reporter: Tilman Hausherr
Assignee: Tilman Hausherr
Fix For: 2.0.0
This code is found in PDFStreamParser
{code}
if (c == '<')
{
COSDictionary pod = parseCOSDictionary();
skipSpaces();
if ((char)pdfSource.peek() == 's')
{
retval = parseCOSStream( pod );
}
else
{
retval = pod;
}
}
{code}
This is incorrect. PDFStreamParser is for content streams. There are no streams
in content streams, the spec requires "All streams shall be indirect objects".
An "indirect object" is something between obj and endobj. But indirect objects
are not allowed in content streams: "Indirect objects and object references
shall not be permitted at all". So parseCOSStream() will never be called. Thus
the new code will be
{code}
if (c == '<')
{
retval = parseCOSDictionary();
}
{code}
To be sure, I tested my own test set and the digitalcopora set (250000 files)
to see whether parseCOSStream is ever called in PDFStreamParser. No it isn't.
How did this incorrect code end up there? Don't know, but it has been there
since 2002.
http://pdfbox.cvs.sourceforge.net/viewvc/pdfbox/pdfbox/src/org/pdfbox/pdfparser/PDFStreamParser.java?revision=1.1&view=markup
Why do I care about this? It is related to a posting in a mailing list by
Andrea Vacondio who mentioned that there are several versions of
parseCOSStream(), so I'm trying to clean up.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@pdfbox.apache.org
For additional commands, e-mail: dev-h...@pdfbox.apache.org
