grobidOrg / grobid / 28080400274

39%

push

github

fix: make block/segmentation desync warning locatable #716 (#1471)

The end DocumentPointer of a document piece can reference a token
position beyond the block's actual token count. The code already
clamps and continues, so the document is still processed, but the
warning named only the block index and token counts -- nothing that
let a user locate the offending region in the source PDF. Reports of
this could therefore never be reproduced.

Enrich the warning with the page number and a short text excerpt of
the block so the next occurrence yields the missing reproduction case,
while keeping the existing clamp behaviour unchanged.

8654 of 24884 branches covered (34.78%)

Branch coverage included in aggregate %.

18430 of 45207 relevant lines covered (40.77%)

4.95 hits per line