Regenerate groundtruth for docling-parse#322 (builtin encoding of embedded symbolic fonts): 16 pages
#4
by wittjeff - opened
Groundtruth for docling-project/docling-parse#322 (builtin encoding of embedded symbolic fonts).
Regenerated with pytest --update-groundtruth tests/test_regression_parse.py on that branch (commit a641751) for the 16 pages whose text changes; every other selected page regenerates byte-identical to the current dataset, so only these files are touched:
- 10017889119412695965-12.pdf pages 2, 4, 5, 9, 11 (HeliosC Type1C: real Cyrillic where the glyph names allow it, Latin-1 transcoding of the cp1251 codes elsewhere, as PyMuPDF reads it)
- 5462902444445925132-4.pdf pages 1-5 (Anastasia TrueType: ASCII back through the post names)
- 10976580690960943929_004.pdf pages 1, 4, 7, 9, 12 (Calibri/Arial symbol subsets: blank glyph reads as a space, Standard-table fabrications become GLYPH markers)
- 4e5aa9238544d683_0009.pdf page 1 (GLYPH<32> reads as a space)
Note: 5462902444445925132-4.pdf pages 1, 3, 4, 5 are also touched by the groundtruth for docling-parse#349 (curly double quotes); whichever lands second needs those four pages regenerated once more on a build carrying both fixes.