No non-printable control characters
Warn on C0 control characters (U+0000–U+001F, excluding tab / LF / CR) and C1 control characters (U+0080–U+009F). These typically arrive from raw paste of terminal output, screen capture, or copy-from-PDF; they render invisibly but break diff / grep / search / JSON encoding.
A frequent source is the literal ESC byte (U+001B) leaked from ANSI color codes when log output is pasted into source.
Binary files are skipped: known binary extensions (images, archives,
fonts, compressed .br/.gz, databases, NumPy/pickle data, office
documents), and any file whose first 8000 bytes hold a NUL byte (the
same rule git uses) or are largely control bytes / invalid UTF-8.
Minified bundles (*.min.js, *.min.css) are skipped too: their
embedded string tables are machine output, not the author’s text.
Test-fixture and test-suite directories (testdata/, fixtures/,
__fixtures__/, __snapshots__/, test/, tests/, __tests__/,
spec/) are skipped: they hold deliberately odd input. Test source
files elsewhere (foo_test.go) are still scanned - write the byte as
an escape (\x1b) instead.
Severity WARN, not BLOCK, because rare text fixtures legitimately contain control bytes.
Fix#
Remove the offending bytes:
LC_ALL=C sed -i 's/[\x00-\x08\x0b\x0c\x0e-\x1f\x7f-\x9f]//g' <file>
Or for the specific ESC byte from ANSI:
sed -i 's/\x1b\[[0-9;]*[A-Za-z]//g' <file>
WARN lets the push through and records the finding. Turn frames on per repo on the dashboard's Policy page - see choosing what the gate checks.
Source on GitHub Live demo How it works Questions: contact@nimblegate.com