tests: resolve through the local dnsmasq cache in CI.

[u5surf]

Fixes the localhost could not be resolved (3: Host not found) failures in t/014-bugs.t TEST 37 and t/058-tcp-socket.t TEST 4, which are red on master today.

Root cause

#2499 added the dnsmasq cache but removed export TEST_NGINX_RESOLVER=8.8.4.4 without putting the cache's address in its place. Nothing sets that variable in CI any more, so each test file falls back to its own default:

$ENV{TEST_NGINX_RESOLVER} ||= '8.8.8.8';

nginx's resolver directive therefore queries a public resolver directly and the cache only ever serves the warm-up digs. Public resolvers do not answer localhost — RFC 6761 leaves that to the stub resolver — hence "Host not found".

Fix

• export TEST_NGINX_RESOLVER=127.0.0.1, so queries actually go through the cache.
• Move dnsmasq to 127.0.0.1:53 and answer the names the resolver tests need with --host-record, since --no-hosts keeps /etc/hosts out.

Port 53 matters: t/087-udp-socket.t TEST 10 hands $TEST_NGINX_RESOLVER to setpeername() with port 53, so the resolver has to stay addressable as a bare IP. systemd-resolved's stub listener owns 127.0.0.53:53 only, and --bind-interfaces keeps dnsmasq off every other address. The step now fails loudly if dnsmasq is not listening there.

TEST 37 and the trailing dot

TEST 37 was added in 5581ded to exercise nginx's resolver against a domain name ending in a dot, so it deserves a name that looks like a real FQDN. It now resolves trailing-dot.test. — multiple labels plus the trailing dot — served from a --host-record on the local dnsmasq, instead of the single-label localhost. it used before. The answer still comes from a real DNS server, just a local one rather than a third party. This follows from the discussion in #2523.

Relationship to #2523

#2523 fixes the same two tests by resolving 127.0.0.1.sslip.io instead of localhost. That trades a dependency on public DNS answering localhost for a dependency on a third-party wildcard DNS service being up and reachable from the runners, which is the direction #2499 moved away from. This patch keeps the resolution local instead.

The rest of #2523 — the --- wait and --max-time adjustments in t/024-access/on-abort.t, t/128-duplex-tcp-socket.t and t/189-http2-subreq-error-wakeup.t — addresses different flakes and does not overlap with this change.

I hereby granted the copyright of the changes in this pull request
to the authors of this lua-nginx-module project.