Repository navigation
chore: release - #4596
chore: release#4596
Conversation
|
Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits. |
dedf0a6 to
e3e6c57
Compare
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: e3e6c57 against main (2b58fcb).
Finding [P3]: rs/moq-tokio/CHANGELOG.md:35 adds a second 0.19.20 section with the same date, compare link, and release notes as the section at line 10. Main already contains this release from #4320; the current PR has no version or implementation changes, only this duplicated history. Remove the duplicate section.
Direction/complexity: the original release update has been superseded by main. Close this redundant PR, or regenerate it only when there is a real release delta.
Verification: inspected the complete current diff and compared the changelog with the pinned base. No build or runtime tests run locally for this documentation-only change. GitHub's Check, Test, and Windows jobs are still running; macOS is queued.
(Written by OpenAI)
e3e6c57 to
205b868
Compare
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: 205b868 against main (5124f81).
No new actionable findings in the regenerated 52-file release delta. All 25 crate versions agree with Cargo.lock and their changelog headings/compare links; the 17 changed workspace dependency requirements match those versions.
The earlier P3 duplicate 0.19.20 changelog finding is resolved. This now releases the Moxygen compatibility work already merged in #4253, so my earlier recommendation to close a redundant PR no longer applies.
Direction/complexity: keep this focused release-plz update. The PR itself adds no implementation, API, or wire changes beyond the pinned base.
Verification limits: reviewed the complete PR diff and release metadata via GitHub; no local build, publish dry-run, runtime, or interop tests run. CI is still queued/running.
(Written by OpenAI)
15e3134 to
e1f22eb
Compare
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: e1f22eb against main (6469f20).
No new actionable findings in the release metadata delta since my previous review. The meaningful refresh is moq-relay 0.17.0 and moq-cli 0.14.0 (rs/moq-relay/Cargo.toml:8, rs/moq-cli/Cargo.toml:8), reflecting the breaking gossip removal already merged in #4601. Both agree with Cargo.lock, the relay workspace requirement (Cargo.toml:159), and their changelog compare links. Rechecked all 25 crate version bumps and 17 changed workspace requirements; they remain consistent.
Direction/complexity: keep this focused release-plz update. Relative to its pinned base, the PR changes only versions, dependency requirements, and changelogs; it adds no implementation, API, or wire changes itself.
Verification limits: static review of all 52 changed-file patches and release metadata via GitHub. No local build, publish dry-run, runtime, or interop tests run. CI is still queued/running.
(Written by OpenAI)
2265aec to
4985030
Compare
|
MERGE — reviewed Pure release-plz churn: package version bumps, CHANGELOG entries, and Checked
Non-blocking
Recommendation: MERGE This is an automated review, not the maintainer's decision |
4985030 to
ad7a03d
Compare
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: ad7a03d against main (eb3e971).
No actionable findings from static review of this release delta. The meaningful change since 4985030 is three third-party lockfile upgrades:
Cargo.lock:3836: lazy_static 1.5.0 → 1.5.1Cargo.lock:6506: pin-utils 0.1.0 → 0.1.1Cargo.lock:7321: rubato 5.0.0 → 5.0.1
Their checksums match the crates.io index, none is yanked, and their registry dependency/feature declarations are unchanged. Rubato still requires Rust 1.87, below this workspace's 1.91 minimum. Rechecked all 25 crate version bumps against Cargo.lock and changelog compare links, plus all 17 changed workspace requirements; they agree.
Direction/complexity: keep the generated release update focused. It adds no MoQ implementation, API, or wire changes relative to the pinned base, but the new dependency versions do change the build inputs, so the previous head's CI results do not validate this head.
Verification limits: inspected all 52 changed-file patches, registry metadata, relevant upstream release notes, and the MoQ resampler integration. No local build, publish dry-run, runtime, or interop tests run; this is not a full dependency-source audit. Android is green; Check, Platform, WASM, Swift, and OBS are still queued.
(Written by OpenAI)
ad7a03d to
830537a
Compare
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: 830537a against main (eb3e971).
No new actionable findings. The meaningful release delta since ad7a03d is moq-auth 0.2.0 → 0.2.1 (rs/moq-auth/Cargo.toml:8), picking up #4600 already merged into the pinned base. The package version agrees with Cargo.lock:4206, the workspace requirement (Cargo.toml:142), and the new changelog heading/compare link (rs/moq-auth/CHANGELOG.md:10). Hang's release notes and moq-room's dependency notes were refreshed accordingly.
Rechecked all 26 crate version bumps and 18 changed workspace requirements against the lockfile and changelog links; they agree. The three third-party upgrades covered by the previous review are unchanged.
Direction/complexity: keep this focused release-plz update. Relative to the pinned base, it adds only release metadata and dependency-version changes, with no new MoQ implementation or wire changes.
Verification limits: static review of the 54-file release diff and comparison with the prior reviewed release commit. No local build, publish dry-run, runtime, or interop tests run. Android is green; Check, Platform, WASM, Swift, and OBS are queued.
(Written by OpenAI)
830537a to
3fbcd79
Compare
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: 3fbcd79, relative to merge base b4234f8 and the prior review at 830537a.
No new actionable findings. The substantive delta is cfg-expr 0.20.9 → 0.20.10 (Cargo.lock:1057) and dns-lookup 4.0.1 → 4.0.2 (Cargo.lock:1973). Both checksums match the crates.io index; neither is yanked. cfg-expr retains its dependencies and Rust 1.85 minimum. dns-lookup lowers its declared minimum to 1.63, scopes cfg-if/socket2 to Unix/Windows, and broadens socket2's range to include the still-locked 0.6.5. No integration mismatch found in moq-tokio's resolver.
All 26 crate versions, 18 changed workspace requirements, lockfile versions, and changelog compare links agree. Other changes are equivalent to the prior reviewed release delta or generated changelog refreshes.
Direction/complexity: keep the release update focused. It adds no MoQ implementation or wire changes relative to its merge base, but the dependency upgrades need this head's CI.
Verification limits: static review of all 54 changed-file patches, registry metadata, relevant upstream changes, and resolver integration. No local build, publish dry-run, runtime, or interop tests; not a full dependency-source audit. Android and Swift are green; Check, Platform, WASM, and OBS are running.
(Written by OpenAI)
9dac434 to
fb6450f
Compare
82f90be to
7ea75cd
Compare
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed 7ea75cd against base 651f199, comparing the release-only delta with the prior review at 3fbcd79.
No new actionable findings. The new build-input change is qmux 0.5.1 → 0.5.2 (Cargo.lock:6868–6871). Its checksum matches the non-yanked crates.io release; dependencies/features are unchanged, and the upstream tagged diff contains the receive-stream race backport. moq-tokio/moq-relay feature wiring remains compatible. The 26 crate versions and 18 changed workspace requirements still agree with the lockfile and changelog headings; the remaining differences are generated release notes for changes already in main and date refreshes.
Direction: keep this focused release update on the compatible 0.5 qmux line. Verification was static metadata/source review, including the upstream qmux diff already checked for #4695. No local build, publish dry-run, runtime, or interop tests run; this is not a full dependency-source audit. CI for this exact head remains queued/running.
7ea75cd to
e846c80
Compare
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed e846c80 against its release parent d76e2c8, comparing the release-only delta with 7ea75cd.
One P2 dependency regression is inline. The four noq/WebTransport 1.3.3 checksums match the non-yanked crates.io releases; features are unchanged, internal requirements align, and Rust 1.88 remains below the workspace's 1.91 minimum. The 26 crate versions and 18 changed workspace requirements are unchanged from the reviewed release. The qmux bump is now in the base.
Direction: the assembler hardening and HTTP/3 close-lifetime fixes are worthwhile, but take a corrected compaction release before shipping this lockfile; keep the security cap. The 1.3.3 source does not include the dev-line raw-QUIC code-mapping change.
Verification: static registry, tagged-source, release-metadata and consumer-wiring review, plus a source-level arithmetic model of the reported case. No Rust build, runtime/interop test or publish dry-run; not a full dependency-source audit.
| [[package]] | ||
| name = "moq-noq-proto" | ||
| version = "1.3.2" | ||
| version = "1.3.3" |
There was a problem hiding this comment.
[P2] Avoid the contiguous-stream rejection in moq-noq-proto 1.3.3
The newly selected assembler can leave more than 1024 buffers after compacting gap-free input. Source-level counterexample: buffer 683 repetitions of 1024/128/128-byte contiguous chunks, each with 16 bytes allocation overhead, without reads. At chunk 2049, min_chunk_size is 854: it retains 683 large chunks and 683 small runs, then rejects 1366 chunks with connection-wide INTERNAL_ERROR. The 874,240-byte backlog fits the default 1,250,000-byte stream window inherited by moq-tokio. Version 1.3.2 had no such rejection. Take a follow-up release that coalesces contiguous runs enough to satisfy the cap, with this mixed-size regression test; don't remove the security limit. This was traced from source, not reproduced in Rust.
6642e70 to
a68febf
Compare
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed a68febf against release parent 4778159, comparing the release-only patch with e846c80.
One new P1 release-compatibility finding is inline. The meaningful delta is moq-net 0.3.9 → 0.4.0 following Fetching's lost unwind-safety traits in #4691. Other dependency selections are unchanged. The prior noq 1.3.3 compaction finding remains unresolved; no duplicate inline comment added.
Direction: avoid turning a main-branch fix into an accidental ecosystem-wide type split. Preserve moq-net's 0.3 compatibility if the trait loss is unintended; otherwise propagate breaking versions through public dependents before release. The generated manifests and lockfile agree internally, but that does not establish downstream compatibility.
Verification: static release-parent comparison, published registry/tag metadata and public API inspection. No build, downstream compile reproduction, publish dry-run or runtime tests run. This head's Android check passed; the remaining build/test checks were still running.
(Written by OpenAI)
| moq-mux = { version = "0.10.9", path = "rs/moq-mux" } | ||
| moq-net = { version = "0.3.8", path = "rs/moq-net" } | ||
| moq-mux = { version = "0.10.10", path = "rs/moq-mux" } | ||
| moq-net = { version = "0.4.0", path = "rs/moq-net" } |
There was a problem hiding this comment.
[P1] Propagate the breaking dependency bump to public dependents
This selects moq-net 0.4 for crates still receiving compatible patch releases, including moq-tokio 0.19.21 and hang 0.21.10. A downstream with moq-net = "0.3.8" and moq-tokio = "0.19.20" can take this patch and then fail at client.with_origin(origin): rs/moq-tokio/src/client.rs:199 now expects the distinct 0.4 origin::Producer, while the caller still builds 0.3. The published 0.19.20 used ^0.3.8. Hang also exposes/re-exports these types. Keep the moq-net dependency compatible or give all affected public dependents breaking version bumps, updating their workspace requirements and release notes. A workspace build with only one moq-net version will not catch this downstream split.
89505b3 to
c1d41d7
Compare
c1d41d7 to
cd59d15
Compare
|
Release cut approved by the maintainer. Merging with a merge commit (release-branch convention). Checked at head
(Written by Claude Opus 5.5) |
🤖 New release
moq-net: 0.3.8 -> 0.3.9 (✓ API compatible changes)moq-json: 0.5.6 -> 0.5.7 (✓ API compatible changes)hang: 0.21.9 -> 0.21.10 (✓ API compatible changes)moq-mux: 0.10.9 -> 0.10.10 (✓ API compatible changes)moq-tokio: 0.19.20 -> 0.19.21 (✓ API compatible changes)moq-video: 0.1.8 -> 0.1.9 (✓ API compatible changes)libmoq: 0.6.9 -> 0.6.10moq-auth: 0.2.0 -> 0.2.1 (✓ API compatible changes)moq-uring: 0.0.10 -> 0.0.11 (✓ API compatible changes)moq-boy: 0.5.9 -> 0.5.10moq-hls: 0.5.9 -> 0.5.10 (✓ API compatible changes)moq-relay: 0.16.0 -> 0.17.0 (⚠ API breaking changes)moq-cli: 0.13.0 -> 0.14.0moq-e2ee: 0.0.9 -> 0.0.10 (✓ API compatible changes)moq-ffi: 0.4.9 -> 0.4.10moq-gst: 0.4.9 -> 0.4.10 (✓ API compatible changes)moq-binary: 0.1.8 -> 0.1.9moq-loc: 0.2.17 -> 0.2.18moq-audio: 0.1.8 -> 0.1.9moq-archive: 0.0.9 -> 0.0.10moq-stats: 0.2.9 -> 0.2.10moq-rtc: 0.3.9 -> 0.3.10moq-rtmp: 0.3.9 -> 0.3.10moq-srt: 0.3.9 -> 0.3.10moq-transcode: 0.1.8 -> 0.1.9moq-room: 0.2.9 -> 0.2.10⚠
moq-relaybreaking changesChangelog
moq-netmoq-jsonhangmoq-muxmoq-tokiomoq-videolibmoqmoq-authmoq-uringmoq-boymoq-hlsmoq-relaymoq-climoq-e2eemoq-ffimoq-gstmoq-binarymoq-locmoq-audiomoq-archivemoq-statsmoq-rtcmoq-rtmpmoq-srtmoq-transcodemoq-roomThis PR was generated with release-plz.