Michael Klink created PDFBOX-4999:
-------------------------------------
Summary: Dangerous COSDictionary.addAll(COSDictionary) method
Key: PDFBOX-4999
URL: https://issues.apache.org/jira/browse/PDFBOX-4999
Project: PDFBox
Issue Type: Bug
Affects Versions: 2.0.21, 3.0.0 PDFBox
Reporter: Michael Klink
The method {{COSDictionary.addAll(COSDictionary)}} creates the impression, by
name and by JavaDoc comment,
{code:java}
/**
* This will add all of the dictionaries keys/values to this dictionary.
...
{code}
that it can be used for exactly that, adding all key/value pairs from the
argument dictionary to the current one, replacing old entries for the same keys.
If one looks at the implementation, though, one is in for a surprise:
{code:java}
/**
* This will add all of the dictionaries keys/values to this dictionary.
* Only called when adding keys to a trailer that already exists.
*
* @param dic The dictionaries to get the keys from.
*/
public void addAll(COSDictionary dic)
{
dic.forEach((key, value) ->
{
/*
* If we're at a second trailer, we have a linearized pdf file, meaning
that the first Size entry represents
* all of the objects so we don't need to grab the second.
*/
if (!COSName.SIZE.equals(key) || !items.containsKey(COSName.SIZE))
{
setItem(key, value);
}
});
}
{code}
Here existing *Size* entries explicitly are not replaced!
This appears to be a relic from times when PDFBox parsed PDF documents front to
back, ignoring cross reference streams, for improved results with linearized
files.
Nowadays this exception does not make any sense anymore, see [this stack
overflow answer|https://stackoverflow.com/a/64502740/1729265]. Furthermore,
this method is used in other contexts than creating trailer unions, even some
PDFBox methods use it to create arbitrary dictionary unions.
Thus, this exception should be removed after all usages of that method in
PDFBox have been analyzed.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
