joule-moq-relay: the loop-of-relays test starts again on new ports when one it chose was taken #652

Merged

dcharlot merged 1 commit from test/relay-loop-port-race into main 2026-11-05 15:12:22 -05:00

Conversation 0 Commits 1 Files changed 1 +27 -12

dcharlot commented 2026-11-05 10:26:40 -05:00

Owner

Copy link

a_loop_of_relays (joule-moq-relay tests/cascade.rs) failed on the linux-arm64 leg of #640 with the relay starts: "binding 127.0.0.1:34891: Address already in use (os error 98)". #640 changes no code here.

Why. The test's three relays each dial the next, so all three ports are chosen with free_port() before any relay listens. free_port() binds port 0, reads the port and lets it go. Another test running at the same time can bind that port in between, and the relay's start is then refused.

The fix. A start refused for a port in use is tried again on three new ports, at most five times. The relays already started are shut down first. Any other refusal still fails the test, and what the test holds of the loop is unchanged:

• B and C hold A's namespace through the right paths;
• it never comes back to A;
• a viewer on C reaches A's publisher;
• a request around the loop is refused with relay_loop at the relay that would close it.

Verified on a Linux build host:

• The retry path runs. A scratch change, not committed, made the first port handed out one a listener keeps. The first start was refused, the test tried again on new ports and passed.
• The cascade tests pass. All 12 passed, three runs in a row.
• No new findings. clippy finds nothing in tests/cascade.rs.

🤖 Generated with Claude Code

`a_loop_of_relays` (joule-moq-relay `tests/cascade.rs`) failed on the linux-arm64 leg of #640 with `the relay starts: "binding 127.0.0.1:34891: Address already in use (os error 98)"`. #640 changes no code here. **Why.** The test's three relays each dial the next, so all three ports are chosen with `free_port()` before any relay listens. `free_port()` binds port 0, reads the port and lets it go. Another test running at the same time can bind that port in between, and the relay's start is then refused. **The fix.** A start refused for a port in use is tried again on three new ports, at most five times. The relays already started are shut down first. Any other refusal still fails the test, and what the test holds of the loop is unchanged: - B and C hold A's namespace through the right paths; - it never comes back to A; - a viewer on C reaches A's publisher; - a request around the loop is refused with `relay_loop` at the relay that would close it. **Verified on a Linux build host:** - **The retry path runs.** A scratch change, not committed, made the first port handed out one a listener keeps. The first start was refused, the test tried again on new ports and passed. - **The cascade tests pass.** All 12 passed, three runs in a row. - **No new findings.** clippy finds nothing in `tests/cascade.rs`. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

dcharlot added 1 commit 2026-11-05 10:26:40 -05:00

joule-moq-relay: the loop-of-relays test starts again on new ports when one it chose was taken ... 7d1be82417

The three relays of a_loop_of_relays each dial the next, so their ports are chosen before any of
them listens, and another test can bind one in between: CI's linux-arm64 leg refused a start
with "Address already in use" (#640). A start refused for a port in use is now tried again on
three new ports, at most five times; any other refusal fails the test, and what the test holds
of the loop is unchanged.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

dcharlot merged commit 55333c2385 into main 2026-11-05 15:12:22 -05:00

dcharlot referenced this pull request from a commit 2026-11-05 15:12:25 -05:00

Merge pull request 'joule-moq-relay: the loop-of-relays test starts again on new ports when one it chose was taken' (#652) from test/relay-loop-port-race into main

dcharlot referenced this pull request from a commit 2026-11-05 16:15:40 -05:00

Merge origin/main (#638, #652: parameter sets, and the relay loop test's ports) into wai/quantum-cloud-tasks

dcharlot referenced this pull request 2026-11-05 16:32:20 -05:00

wai-quantum: receipts say where their counts come from and what was set at the provider, and a second verifier checks them #660

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels

No items

No labels

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

Transaction-Science/open-standards!652

No description provided.