Fix Long Test with IPv6

[maxtropets]

The Debug and Release failures were the same issue: recovery_test timed out in the new recovery_ipv6 subtest.

recovery_ipv6 started correctly, opened an IPv6 network on ::1, then entered the --with-load phase before recovery. It started Locust with an invalid IPv6 URL:

https://::1:<port>

IPv6 literals in URLs must be bracketed:

https://[::1]:<port>

Because the load target URL was malformed, Locust did not drive traffic as expected. The test waited for committed seqnos to advance before beginning recovery, but they never advanced enough, so recovery_test hit the CTest timeout.

Root cause

tests/infra/service_load.py built load target URLs manually:

f"https://{host}:{port}"

That works for IPv4, but not IPv6.

Fix

Use the existing bracket-aware helper:

"https://" + infra.interfaces.make_address(host, port)

This preserves IPv4 output and correctly formats IPv6 as [::1]:port.

Validation

Focused IPv6 recovery test now passes:

CR_FILTER=recovery_ipv6 ./tests.sh --output-on-failure --timeout 600 -R '^recovery_test$'

Result:

100% tests passed, 0 tests failed out of 1

[Copilot]

Pull request overview

This PR fixes an IPv6 regression affecting the long-running load test infrastructure by ensuring the target RPC host URL is constructed with correct bracket handling for IPv6 literals (eg https://[::1]:8000), matching the IPv6 address-formatting helpers introduced in #7671.

Changes:

• Update make_target_host() to build the URL using infra.interfaces.make_address(host, port) so IPv6 hosts are correctly bracketed.
• Add the infra.interfaces import to support the new helper usage.

Custom instructions used:

• .github/copilot-instructions.md

[mergify]

Tick the box to add this pull request to the merge queue (same as @mergifyio queue).

• Queue this pull request