PDFsharp & MigraDoc Foundation

What happens:

http://www.stillhq.com/pdfdb/000419/data.pdf (which is corrupt) causes pdfsharp to hang in an infinite loop.

Cause:

A content stream inside 000419.pdf specifies a length that is longer than the file itself.

After reading a stream, pdfsharp will look for the endstream token. This loop does not exit if the file pointer is at or past the end of the file.

Patch:

See attachment. Patch is against PdfSharp 1.50-beta3b.

_________________
Gerben Vos
Developer

Hi!

Thanks for the submission. Changes look good at first glance.

I will give them a close look on Monday.

_________________
Regards
Thomas Hoevel
PDFsharp Team

We found a few problems while running a huge batch of PDFs from our clients through PdfSharp. For now, these are the most urgent problems we found (having the process hang is pretty terrible). I hope to find some time to look at the remaining problems soon (the LZW decompression bug looks like the worst of those, but there are more).

_________________
Gerben Vos
Developer

In general, I can recommend downloading all of this PDF database and running it through PdfSharp; this should already show up a bunch of problems: http://www.stillhq.com/pdfdb/db.html .

_________________
Gerben Vos
Developer

Thanks for the patch and the link to the PDF DB.

That's an interesting file.
Adobe Reader says the file is corrupt and shows an empty page.
The PDF Reader from Windows 10 shows an empty page without complaining.
The Edge that comes with Windows 10 shows a page full with letter "X" without complaining.
Does that mean that Microsoft included two different PDF viewers in Windows 10 that use different code?

_________________
Regards
Thomas Hoevel
PDFsharp Team