# Binary files must never go through EOL conversion. Without this file, # core.autocrlf=true (a common system-level default on Windows) classifies # small PDFs as text and rewrites their line endings at checkout — in # pdfnative the golden page-tree fixtures grew by 305-616 bytes each while # `git status` stayed clean, which is how the corruption went unnoticed. The # blobs in git were always intact; only the working copy was damaged. # # Text files are LF in the repository AND in every working copy (eol=lf), so # a Windows checkout produces the same bytes as CI and the sample baseline # (tests/regression/baselines/samples.sha256.json) does not depend on the OS. # The two blobs committed as CRLF before this line (both workflow files) were # replaced by the 1.3.0 hardening commit; `verify:docs` (rule eol-lf) fails # on any new one. * text=auto eol=lf *.pdf binary *.icc binary *.ttf binary *.otf binary *.woff binary *.woff2 binary *.png binary *.jpg binary *.jpeg binary *.ico binary # Generated files: collapsed in GitHub diffs and excluded from language stats. .claude/rules/*.md linguist-generated=true tests/regression/baselines/samples.sha256.json linguist-generated=true -diff tests/__snapshots__/*.snap linguist-generated=true package-lock.json linguist-generated=true