fix pdf-hs-driver parsing rejects (or confirm to be valid rejects)

[mtullsen]

Overview

To Do

• [x] low lying fruit is done (generally ERROR-A below) , see 9fba795, 2a103f6, and b964bc0.
• [ ] fix ERROR-B errors
  • [ ] file pdf_tests/2020-03-eval/0377_337cc8e14... points to an "xref stream object" but tries to parse as a traditional xref table! Generally pdf-hs-driver handles "xref streams", so figure out what's different/wrong with this file!
  • [ ] file eval3/**/02ec04d85c4... is a complex file (linearized and hybrid xrefs) and the problem is not apparent yet.
• [ ] a few errors have come to light as a result of the b964bc0 fixes, e.g., pdf_tests/2020-03-eval/0002_00005daa6dd0a1e83744c9e5a2dda68485dd57f2450809923bf7009d89ee1729.pdf
• [ ] see Issue #174 for a whole category of rejects

  Helps

  • Issue #180 would help immensely in completing this one.

Potential errors in the 2020-03-eval corpus

We have a number of variances in the pdf-driver tests. I.e., this test

cd pdf-driver
scripts/get-pdf-tests
scripts/run-tests

Error report from the eval3 corpus

ERROR-A About 70% of your FPs are this ERROR:

INFO:7472 Found xref table.
ERROR:7472 Not enough entries, found UInt 0, but need at least UInt 1

ERROR-B Another 25-ish% is this:

INFO:116 Found xref table.
ERROR:616161 expected "xref"

Those are from files 03/0d/030d6cafff25264297fa3d91dba983d14ed5ba5d69f345078b36c2a0b1c1aa73 and 02/ec/02ec04d85c4a9e65b9693636a8c258875801e6dd1a0dffb8db0d4d124354874a, respectively.

[mtullsen]

The first file, 030d6..., is due to NOT ignoring whitespace pointed to by startxref. the parser should ignore whitespace here.