# codespell ignore list -- read by ci/linter/lint-spelling.sh when present.
#
# EMPTY ON PURPOSE, and shipped anyway so the seam is documented rather than
# discovered. codespell's dictionary flags plenty of real identifiers, and an
# adopter meets that on the first run: nginx-cache-turbo-module measured 151
# findings, 149 of them correct spellings in that tree -- its own domain
# acronym, an abbreviation for a well-known HTTP header, a URL path segment,
# another header name, and bytes inside a fuzz corpus input.
#
# Without a file to put those in, the options are a red gate on arrival or a
# blanket suppression, and the second one retires one of the cheapest real
# checks in the set.
#
# Add one word per line, lowercase, with a comment saying why it is correct
# HERE. An entry earns its place by being WRONG for this tree, never by being
# inconvenient -- a word that is genuinely misspelled gets fixed instead.
#
# Note the file is itself linted, so spell out an example rather than quoting
# the flagged token: writing the word here is what puts it in the tree.
#
# Corpus files are byte-exact fuzz inputs and must never be "corrected" --
# editing one silently changes what the fuzzer replays. lib.sh already excludes
# ci/fuzz/corpus/, so those need no entry here.

# Token-end index in the Accept-Encoding scanner, paired with the start index
# two letters earlier in the alphabet. Both are local size_t cursors in
# ci/fuzz/fuzz_accept_encoding.c, not an abbreviation of "the".
te

# nginx's own field name on ngx_http_request_t -- the filename extension of the
# request URI. Spelled without the final "sion" by nginx itself since 0.1, so
# correcting it would stop the code compiling.
exten
