v0.23.0

Breaking changes

• ParserError::OversizedRecord { length }: an MRT header declaring a body over the 16 MiB limit now returns this variant instead of Unsupported("MRT message too large: …"). The Display text is unchanged. Downstream exhaustive matches on ParserError need a new arm.
• ParserBmpError carries the failing value and is #[non_exhaustive]: From<io::Error> and the unknown-message-type conversion collapsed every failure into InvalidOpenBmpHeader and dropped the cause; they now return the new IoError { kind, message } and UnknownMessageType(u8) variants. Downstream matches on the enum now need a wildcard arm: #[non_exhaustive] means the two new variants cannot be covered by adding arms alone.

Added

• BGP Domain Path (D-PATH) attribute support (RFC 10039): path attribute type 36 now parses into AttributeValue::DomainPath(DomainPathAttribute) — segments of DomainPathDomain { global_admin, local_admin, isf_safi_type } — and re-encodes byte-identically instead of falling into the Unknown raw catch-all. Malformed values, and a D-PATH on a family other than IPVPN or EVPN, become RFC 7606 validation findings with the bytes retained raw, and the encoder refuses values it cannot represent. Text rendering and byte-level dissection label the attribute BGP_DOMAIN_PATH.
• --color auto|always|never for --format text: colors session keys, section headers, prefixes, and next-hop values with ANSI accents that follow the terminal theme. auto (the default) colors only when stdout is a terminal; NO_COLOR disables coloring and CLICOLOR_FORCE forces it. Library callers opt in with render::text::Style::ansi() plus the new format_record_with_style / format_record_with_hex_and_style; format_record and the unstyled output are unchanged, since styling is a post-pass over the rendered block.
• parse_ris_live_message_iter: parses one RIS Live JSON message into an iterator over its elems instead of a Vec. Every elem carries its own copy of the message's AS path and communities, so collecting a message that announces many prefixes needs memory proportional to prefixes × path length; the iterator makes those copies one elem at a time. parse_ris_live_message is unchanged and now builds on it.

Changed

• Internal cleanups, no behavior change: the MP next-hop encoder (encode_mp_next_hop) is shared by the NEXT_HOP and MP_REACH encoders instead of being duplicated, AsPathSegment hashing skips the sort when a set is already ordered, Elementor::record_to_elems logs peer-table conversion errors as its documentation promises, the invariant unreachable!() arms state their invariant, and both the crate-wide uninlined_format_args allow and a module-wide #![allow(unused)] are removed.
• --format text session labels are now PEER/LOCAL: the endpoint lines read PEER: <peer_ip> AS<peer_asn> and LOCAL: <local_ip> AS<local_asn>, matching the peer/local names MRT (RFC 6396) uses for the same fields instead of the ambiguous FROM/TO. The rendered values are unchanged.
• Record parse errors are logged once per parser: the default, raw, update and route iterators log the first ParseError of a parser at error level and later ones at debug level. A corrupted stream re-synchronises header by header, which previously emitted one error line per 12 bytes of damaged input. Other error classes are logged as before.
• Faster elem conversion: into_elem_iter is about 24% faster on updates and 21% faster on RIB dumps, mostly from fewer copies and allocations when building elems; output is unchanged.

Fixed

• Iterators stay aligned after an unknown MRT entry type: chunk_mrt_record rejected an unknown entry type after reading only the 12-byte header, so into_record_iter, into_elem_iter, into_raw_record_iter and the fallible iterators read the record body as the next header and misparsed or dropped every record after it. The body is now consumed first, and the error's bytes hold the whole record. into_diagnostic_iter reports the record as a ParseError event and continues, where it previously stopped at any framing error.
• Iterators stop at an oversized MRT length: after a header declaring a body over 16 MiB, the default and fallible iterators read that body as headers. The default iterators now log the new OversizedRecord error and end, and the fallible iterators yield it once and end. into_recovering_record_iter still scans forward.
• RIS Live JSON elems carry origin_asns: parse_ris_live_message left origin_asns empty, so an origin_asn filter never matched an elem parsed from RIS Live's JSON fields. It is now derived from the AS path, as parse_ris_live_message_raw and the MRT iterators already did.
• Dissection span correlation is linear in the number of warnings: span_record_warnings (used by DiagnosticIterator::with_dissection) walked the whole dissection tree once per warning, so a BGP UPDATE carrying thousands of duplicate attributes took tens of seconds to diagnose. The tree is now indexed by field name once per record. Spans are unchanged.
• Fallible iterators stop after a fatal stream read error: into_fallible_record_iter, into_fallible_elem_iter, into_fallible_update_iter and into_fallible_route_iter now yield a framing I/O or decompression error once and then end, instead of polling a spent decoder indefinitely. An Interrupted/WouldBlock read stays retryable only when it consumed no framing bytes, and skippable malformed records are unchanged (#350).

Contributors

• @ties — faster elem conversion (#348), parse_ris_live_message_iter and RIS Live origin ASNs, once-per-parser parse error logging, linear-time dissection span correlation, MRT framing after unknown entry types and oversized lengths