# The corpus is data, not source, and its bytes are the measurement.
#
# Git's newline conversion is helpful for code and destructive here. Checked out
# on Windows with the default `core.autocrlf`, every `\n` in a sample becomes
# `\r\n` -- and these samples exist precisely to hold line endings nobody else
# counts the same way: a NEL from an EBCDIC conversion, a form feed, a progress
# bar that redraws with carriage returns. Convert them and the corpus stops
# being a record of what tools actually print and becomes a record of what git
# thought it should print.
#
# Measured: on windows-latest the `cargo-de` sample arrived with CRLF and the
# test that proves `str.splitlines()` would miscount it found nothing to prove.
test/korpus/** -text
test/korpus-kaynak/** -text

# Source keeps LF in the repository whatever the checkout does, so a diff never
# turns into every line at once.
*.py text eol=lf
*.md text eol=lf
*.toml text eol=lf
*.yml text eol=lf
