You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The 0.11.0 release PR (#884) fails 17 checks: Bundle resolution, every server image build, CPU containers, both CUDA 13 compatibility jobs, and Release candidate / Complete. Every one fails with bundle lock inputs changed; recompile and review the complete closure.
Each saved bundle manifest binds the SHA256 of the whole root uv.lock. The release PR refresh runs uv lock, which rewrites the workspace package versions (0.10.1 -> 0.11.0). That makes all 16 manifests stale, even though no input of any bundle closure changed. 0.11.0 is the first release since the saved bundle locks landed in #802.
This change binds each manifest to the exact third-party versions in the root lock instead of to the file's bytes. The input digest is renamed from uv_lock_sha256 to root_versions_sha256. All 16 manifests are regenerated with the existing 2026-10-18T00:00:00Z cutoff. No pylock.toml changes.
Why not recompile in the release workflow
The refresh step runs in the protected release-automation job, which holds the release App's write token. It deliberately runs no third-party code (--ignore-scripts). A compile there would need a full root environment sync and uv 0.9.28. The resolve would also have to build metadata for 10 sdist-only distributions and one Git source, so third-party build code would run next to the write token.
The release contract checker pins that step to an exact command sequence.
The compile would only re-stamp a hash. See evidence item 1.
Evidence
I recompiled all 16 targets against the current release branch head (ff398c2) with the existing cutoff and uv 0.9.28. All 16 pylock.toml files stay byte-identical. The only change in each manifest is uv_lock_sha256.
The release branch's root lock has the same third-party name/version set as main (370 distributions). The release job's uv 0.5.31 also drops the graalpy wheels of jiter, pybase64 and pydantic-core. No bundle closure uses those wheels, so the exception check still passes.
On this branch, recompiling all 16 targets with the same cutoff leaves every pylock.toml byte-identical. Each manifest changes in two places only: policy_sha256, because the verifier and compiler sources changed, and the renamed input digest.
tools/ci/bundle_lock.py check passes on this branch. It also passes, without a recompile, on the release-branch tree with this change applied.
What stays bound
Every third-party root distribution and its exact versions, including local labels such as +cu129. Any root version change still invalidates every manifest.
Artifact sources. check_lock already reruns validate_exceptions against the current root lock on every check. A new test shows that substituting an artifact at the same version still fails.
Unchanged: bundle declarations, target, uv version, exceptions, source pins, lock hash, and the compiler, verifier and declaration-resolver source hashes.
Workspace members and lock serialization are no longer bound. They never reach bundle-libs: root_packages already skips editable and virtual workspace entries when it derives constraints.
Release flow
This needs no change to the workflow or the release contract. After it merges, the next release-PR refresh rewrites uv.lock without invalidating the saved locks.
This changes the bundle-lock integrity binding from #802, so it needs review from the bundle-lock owner.
Each saved bundle manifest bound the SHA256 of the whole root uv.lock.
A release PR refresh rewrites the workspace package versions in that
file, so every manifest went stale and the verifier rejected all 16
targets with "bundle lock inputs changed" although no closure input
changed. Recompiling the 16 targets against the 0.11.0 release branch
with the existing 2026-10-18T00:00:00Z cutoff leaves every pylock.toml
byte-identical; only the root lock hash moves.
Bind the manifest to the exact third-party name/version set of the root
lock, which is what compilation constrains. Workspace members and lock
serialization no longer invalidate a closure. Artifact sources remain
checked on every verification through the exception policy, and any
root version change still requires recompilation.
Regenerate all 16 manifests with the same cutoff and uv 0.9.28. No
pylock.toml changes; manifests change only in policy_sha256 and the
renamed root_versions_sha256 input.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
No actionable comments were generated in the recent review.
Walkthrough
Bundle lock manifests now bind to exact third-party root package versions rather than the serialized root lock file. The checker and CI compiler use the new digest. Tests cover ignored lock metadata, root dependency changes, and artifact hash changes.
Changes: Bundle lock binding
Layer / File(s)
Summary
Root-version digest and verification packages/sie_server/src/sie_server/bundle_lock.py, tools/ci/bundle_lock.py, packages/sie_server/tests/core/test_bundle_lock.py
The checker and CI compiler use a digest of canonical third-party package names and sorted versions. Tests cover workspace and serialization changes, dependency-set changes, and artifact hash changes.
Saved manifests replace uv_lock_sha256 with root_versions_sha256 and update policy_sha256; lock_sha256 remains unchanged. The README describes the revised binding inputs.
Possibly related PRs
superlinked/sie#855: Its generated manifests use uv_lock_sha256, which this change replaces with root_versions_sha256.
superlinked/sie#878: It updates Transformers 5 bundle locks and manifests that this change also revises.
superlinked/sie#845: It adds the saved embedding manifest included in these manifest updates.
Suggested reviewers:svonava, krisztian-gajdar
Merge Risk: Minimal · up to 7736b
The new binding is consistent and no merge-blocking risk is established.
Pre-merge checks: 4 · 1
Failed checks: 1 warning
Check name
Status
Explanation
Resolution
Docstring Coverage
Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documenta…
Write docstrings for the functions missing them to satisfy the coverage threshold.
Passed checks: 4 passed
Check name
Status
Explanation
Description Check
Check skipped - CodeRabbit’s high-level summary is enabled.
Title check
The title "fix(bundles): bind saved bundle locks to third-party root versions" matches the main change. The PR replaces the full uv.lock hash (uv_lock_sha256) with root_versions_sha256, a digest…
Linked Issues check
Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check
Check skipped because no linked issues were found for this pull request.
Full details: Title check
Explanation
The title "fix(bundles): bind saved bundle locks to third-party root versions" matches the main change. The PR replaces the full uv.lock hash (uv_lock_sha256) with root_versions_sha256, a digest of exact third-party root package versions. The 16 bundle-lock manifests, bundle_lock.py, the tests, and the CI script all implement this change. The title names the bound input (third-party root versions) and the subject (saved bundle locks). It does not name the rename or the policy_sha256 update, which are secondary details.
Full details: Docstring Coverage
Explanation
Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentation-only edits and deletions within surviving functions are included; unchanged diff context and fully deleted functions are excluded. Analyzed 4 functions across 3 files. (16 skipped: 16 unsupported.)
Resolution
Write docstrings for the functions missing them to satisfy the coverage threshold.
Finishing Touches
Generate docstrings
Commit to this branch
Create a new PR
Generate unit tests (beta)
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
CodeRabbit Verify
Run CodeRabbit Verify to test the changes in this pull request.
Recent review info
Run configuration
Source mode: Auto
Repository: Unavailable
Branch: Unavailable
File: No configuration file from this repository applied
Commit: Unavailable
Layers: Organization UI
Review profile: CHILL
Plan: Team
Run ID: e8169f54-0f37-46aa-b167-66d9d10bedd1
Commits
Reviewing files that changed from the base of the PR and between aa27832 and 7736b91.
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 6 remain after this review.
The root lock records torchvision 0.24.1 from two registries. Hashing
the version list made the binding depend on how many source rows share
a version, so dropping or adding one source row would invalidate all 16
manifests although the compilation constraints stay the same. Hash the
set of exact versions; artifact sources stay checked on every
verification through the exception policy.
Regenerate all 16 manifests with the 2026-10-18T00:00:00Z cutoff and
uv 0.9.28. No pylock.toml changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The reason will be displayed to describe this comment to others. Learn more.
🔵 Needs a closer look
Bundle-lock owner review is needed to confirm that narrowing integrity checks from all root-lock bytes to version-only binding plus exception-policy checks preserves supply-chain guarantees across supported targets.
No open issues 🎉
What changed in this PR
This PR prevents release-only workspace version bumps from invalidating SIE’s saved bundle locks, addressing the release failure described in #884. It replaces the whole-file root-lock digest introduced in #802 with a digest of exact third-party package versions.
Changes:
Update compilation and verification to use root_versions_sha256, while retaining artifact validation through the exception policy.
Add regression tests for tolerated workspace and serialization changes and rejected artifact substitutions.
Refresh all 16 manifests and document the revised binding; no pylock.toml files change.
File
Description
tools/ci/bundle_lock.py
Write the third-party version digest during compilation.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Summary
The 0.11.0 release PR (#884) fails 17 checks: Bundle resolution, every server image build, CPU containers, both CUDA 13 compatibility jobs, and Release candidate / Complete. Every one fails with
bundle lock inputs changed; recompile and review the complete closure.Each saved bundle manifest binds the SHA256 of the whole root
uv.lock. The release PR refresh runsuv lock, which rewrites the workspace package versions (0.10.1 -> 0.11.0). That makes all 16 manifests stale, even though no input of any bundle closure changed. 0.11.0 is the first release since the saved bundle locks landed in #802.This change binds each manifest to the exact third-party versions in the root lock instead of to the file's bytes. The input digest is renamed from
uv_lock_sha256toroot_versions_sha256. All 16 manifests are regenerated with the existing2026-10-18T00:00:00Zcutoff. Nopylock.tomlchanges.Why not recompile in the release workflow
release-automationjob, which holds the release App's write token. It deliberately runs no third-party code (--ignore-scripts). A compile there would need a full root environment sync and uv 0.9.28. The resolve would also have to build metadata for 10 sdist-only distributions and one Git source, so third-party build code would run next to the write token.Evidence
ff398c2) with the existing cutoff and uv 0.9.28. All 16pylock.tomlfiles stay byte-identical. The only change in each manifest isuv_lock_sha256.main(370 distributions). The release job's uv 0.5.31 also drops the graalpy wheels ofjiter,pybase64andpydantic-core. No bundle closure uses those wheels, so the exception check still passes.pylock.tomlbyte-identical. Each manifest changes in two places only:policy_sha256, because the verifier and compiler sources changed, and the renamed input digest.tools/ci/bundle_lock.py checkpasses on this branch. It also passes, without a recompile, on the release-branch tree with this change applied.What stays bound
+cu129. Any root version change still invalidates every manifest.check_lockalready rerunsvalidate_exceptionsagainst the current root lock on every check. A new test shows that substituting an artifact at the same version still fails.Workspace members and lock serialization are no longer bound. They never reach
bundle-libs:root_packagesalready skips editable and virtual workspace entries when it derives constraints.Release flow
This needs no change to the workflow or the release contract. After it merges, the next release-PR refresh rewrites
uv.lockwithout invalidating the saved locks.This changes the bundle-lock integrity binding from #802, so it needs review from the bundle-lock owner.
Test plan
pytest packages/sie_server/tests/core/test_bundle_lock.pypython tools/ci/bundle_lock.py check(this branch, and the release-branch tree with this change)SIE_RUN_IMAGE_RESOLUTION=1 pytest packages/sie_server/tests/core/test_bundle_image_resolution.pywith uv 0.9.28python tools/ci/check_release_contract.pymise run lint,mise run typecheck🤖 Generated with Claude Code
Summary by CodeRabbit