[
https://issues.apache.org/jira/browse/PDFBOX-4750?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17023527#comment-17023527
]
Tilman Hausherr commented on PDFBOX-4750:
-----------------------------------------
Snapshot available at
https://repository.apache.org/content/groups/snapshots/org/apache/pdfbox/pdfbox-app/2.0.19-SNAPSHOT/
I also did a test on the attached file, with this code:
{code}
try (PDDocument doc = PDFParser.load(new File("XXXX/PDFBOX-4750.pdf")))
{
PDFRenderer r = new PDFRenderer(doc);
for (int i = 0; i < doc.getNumberOfPages(); ++i)
{
BufferedImage bim1 = r.renderImageWithDPI(i, 96);
ImageIO.write(bim1, "png", new File("XXXX/in/PDFBOX-4750-saved.pdf-" +
(i+1) + ".png"));
PDPage page = doc.getPage(i);
PDStream newContent = new PDStream(doc);
try (InputStream is = page.getContents();
OutputStream os =
newContent.createOutputStream(COSName.FLATE_DECODE))
{
PDFStreamParser parser = new PDFStreamParser(is);
parser.parse();
ContentStreamWriter tokenWriter = new ContentStreamWriter(os);
tokenWriter.writeTokens(parser.getTokens());
}
page.setContents(newContent);
}
doc.save(new File("XXXX/PDFBOX-4750-saved.pdf"));
}
TestPDFToImage testPDFToImage = new
TestPDFToImage(TestPDFToImage.class.getName());
if (!testPDFToImage.doTestFile(new File("XXXX/PDFBOX-4750-saved.pdf"),
"XXXX/in", "XXXX/out"))
{
fail("Rendering failed or is not identical");
}
{code}
and it succeeded. It also succeeded after the first commit. The following ones
are for clarity and to prevent having too many unneeded spaces in the modified
content stream. You should use the test code above to create your own test. It
creates a first rendering set in the "in" directory and then modifies the file
and then compares the renderings.
> java.io.IOException: Error:Unknown type in content stream:COSNull{}
> -------------------------------------------------------------------
>
> Key: PDFBOX-4750
> URL: https://issues.apache.org/jira/browse/PDFBOX-4750
> Project: PDFBox
> Issue Type: Bug
> Components: Writing
> Affects Versions: 2.0.8, 2.0.18
> Reporter: tomas kochan
> Assignee: Tilman Hausherr
> Priority: Major
> Fix For: 2.0.19, 3.0.0 PDFBox
>
> Attachments: 01 - K17 - Was dahinter steckt - dsb.pdf,
> contentAllOperatorsOfCorruptedPage.txt
>
>
> By removing some optional content for specific document, which is bordered
> with Operator BDC and EMC, we are facing an issue by writing the changed set
> of tokens into PDStream.
> The code looks like:
> PDStream updatedStream = new PDStream(document);
> OutputStream out = updatedStream.getCOSObject().createRawOutputStream();
> ContentStreamWriter tokenWriter = new ContentStreamWriter(out);
> tokenWriter.writeTokens(result);
> out.flush();
> out.close();
> page.setContents(updatedStream);
>
> The following exception occurs at line 'tokenWriter.writeTokens(result);' :
> java.io.IOException: Error:Unknown type in content stream:COSNull{}
> at
> org.apache.pdfbox.pdfwriter.ContentStreamWriter.writeObject(ContentStreamWriter.java:199)
> at
> org.apache.pdfbox.pdfwriter.ContentStreamWriter.writeObject(ContentStreamWriter.java:146)
> at
> org.apache.pdfbox.pdfwriter.ContentStreamWriter.writeObject(ContentStreamWriter.java:181)
> at
> org.apache.pdfbox.pdfwriter.ContentStreamWriter.writeTokens(ContentStreamWriter.java:109)
> at
> de.justiz.eip.pdf.tools.PdfContext.getOrRemoveOptionalTextContentfromPage(PdfContext.java:429)
> at
> de.justiz.eip.pdf.tools.paging.PagingInfoInterpreterPdfContext.removePagingInfo(PagingInfoInterpreterPdfContext.java:325)
>
> After the analyze we figured out two issues:
> 1. We assume, the Pdf Document it's self is corrupted, It contains on some
> place operator BI, which is based on the PDF-Reference-V1.7 a begin of inline
> image object. This Operator is not followed by "ID" or "EI" operator.
> Extract from list of Tokens:
> next PDFOperator\{Do}
> next COSFloat\{0.016674607}
> next COSInt\{0}
> next COSInt\{0}
> next COSFloat\{0.061831153}
> next COSFloat\{0.070509767}
> next COSFloat\{-0.302021403}
> next PDFOperator\{cm}
> next PDFOperator\{BI}
> next PDFOperator\{Q}
> next PDFOperator\{Q}
> next COSName\{OC}
> next COSName\{eAkteOptionalContent7}
> next PDFOperator\{BDC}
> Moreover one "DP" Entry in the "BI" operator's COSDictionary contains
> COSArray with COSNull values. However the assumption is, that the COSNull
> values are not forbidden in the Pdf content.
>
> COSDictionary\{COSName{Interpolate}:true;COSName\{W}:COSInt\{35};COSName\{H}:COSInt\{26};COSName\{CS}:COSName\{RGB};COSName\{BPC}:COSInt\{8};COSName\{F}:COSArray\{[COSName{A85},COSName\{DCT}]};COSName\{DP}:COSArray{[COSNull{},COSNull{}]};}
> 2. Despite wrong content in the pdf-document (described above) the PDF-Box
> api crashed by storing this operators into PDStream by his inability to
> recognize COSNull in the method
> org.apache.pdfbox.pdfwriter.ContentStreamWriter.writeObject(Object)
>
> The assumption on this place is, that the method "writeObject" forgot to
> cover COSNull as an valid input. The org.apache.pdfbox.cos.COSNull.NULL is
> valid Object, which is broadly used by PDF-Api itself.
> The Method
> org.apache.pdfbox.pdfwriter.ContentStreamWriter.writeObject(Object) PDF-Api
> 2.0.8, also in 2.0.18 doesn't cover the COSNull case in it's if/else
> conditions, instead of it throws the new IOException( "Error:Unknown type in
> content stream:" + o ).
>
> Could you confirm, that the method writeObject contains bug and should be
> corrected to cover also COSNull Object? If so, in which version could we
> expect the fix?
> Thank you
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
