parseCSV skipped a record on any read error and asked for the next one.
When the download of as.csv broke off, the response body returned the
same error on every read, so the generator logged it forever and make
asupdate never finished.
parseCSV now skips only a record that does not parse, a csv.ParseError.
Any other read error ends it with that error, so the generator exits
non-zero and script/asupdate keeps the old list.
Model: opus-5-5
make asupdate piped the generator into gzip, so the target saw gzip's
exit status. A failed download replaced pkg/asinfo/asdata.json.gz with
an empty file and the target still exited 0.
The target now runs the new script/asupdate. It writes the output of
cmd/asinfo-gen to a temporary file. Only when the generator succeeds
does it compress the list and move it into place. Otherwise it leaves
the file alone and exits non-zero. The generator now fails when as.csv
holds no AS.
The generator reads as.csv from ipverse/as-metadata, the source's new
name. The list it builds there is unchanged, so no new list is
committed.
Model: opus-5-5
The built-in AS list had no letter outside ASCII, and make asupdate
rejected today's as.csv, which added a country code column.
The generator now checks only the first three header columns and reads
past the rest; the parsing moved into parseCSV. The list is regenerated.
The new source spells some names differently and no longer lists AS 0
or the documentation AS numbers, so the tests that relied on those
entries changed.
A WHOIS answer that is not valid UTF-8 is now decoded as Latin-1, which
RIPE and APNIC send, before it is parsed and stored. LACNIC answers in
UTF-8.
Model: opus-5-5