# assertion-audit baseline -- a RATCHET, not a clean bill of health.
#
# Every line is one OCCURRENCE of a finding that already existed when the gate
# was introduced, not one unique finding TEXT. The gate blocks NEW findings;
# these are known debt. Deleting a line is always safe (the gate simply starts
# enforcing that one occurrence again). Adding one requires the same
# justification as an inline `# audit: allow-<code> <why>` pragma.
#
# Format: <path>::<rule>::<fingerprint>  # offending source
#
# Several DIFFERENT lines in the same file can have byte-for-byte identical
# source (e.g. six copies of `assert out.error is not None` in a
# table-driven test) and therefore the identical fingerprint -- the
# fingerprint is deliberately content-only, not line-based, so a baseline
# entry survives the statement moving when unrelated code above it changes
# (see Finding.fingerprint). When more than one occurrence in a file shares a
# key, each occurrence gets its own line here, disambiguated with a trailing
# `@N` (e.g. `...::548cd5c7b83b@1` .. `...::548cd5c7b83b@6`). The `@N` suffix
# carries no meaning beyond "make this line's text distinct from its
# siblings so a plain set doesn't collapse six occurrences into one" --
# matching counts occurrences per bare key, so deleting ANY one `@N` line
# (or renumbering them) has the same effect: it lowers that key's suppressed
# count by exactly one. A key with only one occurrence in the file is never
# suffixed, so the vast majority of entries are unchanged from before.

