Dumping objects when the xref table is corrupted

[davidtr1037]

Currently, if the xref table in the PDF is corrupted (wrong offsets), then dumppdf.py fails to extract files.

For example, when the following command runs:

dumppdf.py -E /tmp/ sample.pdf

We get:

Traceback (most recent call last):
  File "/usr/local/bin/dumppdf.py", line 268, in <module>
    if __name__ == '__main__': sys.exit(main(sys.argv))
  File "/usr/local/bin/dumppdf.py", line 265, in main
    dumpall=dumpall, codec=codec, extractdir=extractdir)
  File "/usr/local/bin/dumppdf.py", line 191, in extractembedded
    obj = doc.getobj(objid)
  File "/usr/local/lib/python2.7/dist-packages/pdfminer/pdfdocument.py", line 681, in getobj
    stream = stream_value(self.getobj(strmid))
  File "/usr/local/lib/python2.7/dist-packages/pdfminer/pdfdocument.py", line 694, in getobj
    raise PDFObjectNotFound(objid)
pdfminer.pdftypes.PDFObjectNotFound: 1

Possible solution:
When computing an object offset (in PDFXref.load), you can check if the offset given by the xref table makes sense, probably by checking if starts with "<objid> 0 obj". If that's not the case, then you can find that pattern in the whole file, and use that offset instead.

[RolandColored]

Maybe it's possible to fix the corrupt PDF using a different tool beforehand. I'm asking if it makes sense to start dealing with all possible edge cases or just assume valid PDF files.