Possible race condition in zone_cache:put_zone/1?

[g-andrade]

Hi,

I've been using erldns 2.2.0 as a library (great library, by the way!)

I happened to take a look into erldns_zone_cache:put_zone/1 and, if the cache works the way I superficially understood it to, there appears to be race condition between invoking delete_zone_records/1 and put_zone/2.

erldns/src/erldns_zone_cache.erl

Lines 277 to 283 in 97d2b8e

	put_zone({Name, Sha, Records, Keys}) ->
	SignedZone = sign_zone(build_zone(Name, Sha, Records, Keys)),
	NamedRecords = build_named_index(SignedZone#zone.records),
	delete_zone_records(erldns:normalize_name(Name)),
	case put_zone(erldns:normalize_name(Name), SignedZone#zone{records = trimmed}) of
	ok ->
	put_zone_records(erldns:normalize_name(Name), NamedRecords);

Did I understood it correctly? Is there a tiny window where lookups might fail every time put_zone/1 is used?
If so, would you be open to changing the order between calls? I'd be available for a PR.

Thanks!

[g-andrade]

(Bump)

[g-andrade]

(Bump 🥺)