fix(rs): repair the test call sites the main merge missed - #3287
fix(rs): repair the test call sites the main merge missed#3287kixelated wants to merge 1 commit into
Conversation
The main-into-dev merge (3227c19) left three tests behind, and the first two stopped the workspace compiling, so `just check` and `just test` failed for every branch based on dev. `moq-video` called `catalog.track_info()` in a test while dev's `moq-mux` signature takes a priority; the production call site in the same file was already updated. `moq-rtc` built its origins with `moq_net::Origin::random()` and announced through `create_broadcast(path, Route)`, both replaced on dev by `origin::Info` plus a separate `announce(prefix, Route)`. `moq-video`'s idle-capture test then compiled and failed. The discontinuity marker is still published; the test read the finished track through the default `Duration::ZERO` budget, which sheds the empty group as soon as the resumed group stamps its reach. A live consumer takes that marker as the live edge, so the fix is a subscription with a replay window, matching the other tests that walk a finished multi-group track. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Closing: every breakage this PR repaired landed first, so nothing here is still needed.
The one thing still outstanding is unrelated to the build: (Written by claude-opus-5) |
1 participant
Summary
origin/devat 3227c19 ("Merge main into dev") does not compile: the merge carried three tests over frommainwithout updating them for dev's APIs. The first two are hard errors, and both crates are dependents of widely-used ones, sojust checkandjust testfail for essentially every branch based ondev(that is how I ran into this, from #3282).moq-video(encode/producer.rs): a test calledcatalog.track_info(), but dev'smoq_mux::catalog::Producer::track_infotakes apriority: u8(feat(net)!: divide one connection's bandwidth estimate among its tracks #2854). The production call site in the same file was updated by the merge; this one was missed.moq-rtc(lib.rs): the WHIP/WHEP round-trip test built origins withmoq_net::Origin::random().produce()and announced by passing aRoutetocreate_broadcast. On dev an origin isorigin::Infoplus a driver (moq_tokio::origin::spawn, already a dev-dependency), and announcing is separate from creating:create_broadcast(path)thenannounce(prefix, Route), in create-populate-announce order. The announcement assertions move frompath/broadcast.is_some()toprefix/active.moq-videoagain: with the first fix in place,idle_capture_publishes_a_discontinuity_before_resumecompiled and failed ([1, 1], expected[1, 0, 1]) — a real behavior difference the compile error had been hiding.Root cause: the marker is still published (
container::Producer::discontinuityappends an empty group and finishes it). What changed is delivery. A group's staleness is measured by its reach, which is bounded by its successor's start, so the empty group's reach is the resumed group's timestamp — equal to the live edge, and>=aDuration::ZERObudget. The default subscription therefore sheds it. That is correct for a live consumer, which takes the marker as the live edge when it arrives with no successor; it is wrong only for a test that walks the finished track afterwards. Fixed by subscribing with a replay window, the same thing themoq-binary/moq-jsonstream tests do for the same reason.Test-only changes; no production code touched.
Public API changes
None.
Test plan
just checkandjust test(diff-scoped tomoq-rtc,moq-video, and their dependents): 376/376 pass. Both failed onorigin/devbefore this.cargo test -p moq-rtc: 44 pass, includingwhip_and_whep_round_trip_opus, which had not been able to run.just check-all/just test allis running; I will note anything else it turns up.Not fixed here
just drafts checkfails ondraft-lcurley-moq-hang.md:425, a{{field-timeline}}cross-reference with no anchor (from #3109). Thetimelinefield exists on thejsonandbinaryrendition schemas, but the draft never got a### timeline {#field-timeline}section, and the sentence claims the field "carries the same meaning here as for a media rendition" when media renditions have no such field. Writing that section is spec authoring rather than a build repair, so it wants an author. Nothing in CI runs the drafts check, so it blocks no workflow.(Written by claude-opus-5)