# Golden files are compared BYTE-for-byte by the test suite
# (tests/symbol_import/run_golden.py uses filecmp with shallow=False), and the
# importer that produces them writes LF explicitly (newline="\n"). A checkout
# that translates line endings therefore rewrites the goldens to CRLF while the
# produced files stay LF, and the compare fails — on Windows, where
# core.autocrlf defaults to true, symbol_import_golden fails for this reason
# alone. Worse, the failure is near-silent: the test's diff uses splitlines(),
# which normalises line endings, so it reports the files as differing while
# printing no differences at all.
#
# Pin them so the bytes on disk match the bytes committed, whatever the
# consumer's git config.
tests/symbol_import/golden/* -text
