feat(installer): publish binary and assets as one atomic bundle - #268
Open
Alan-TheGentleman wants to merge 4 commits into
Open
feat(installer): publish binary and assets as one atomic bundle#268Alan-TheGentleman wants to merge 4 commits into
Alan-TheGentleman wants to merge 4 commits into
Conversation
…le atomically Extend the installer's binary-only bundle into one atomic binary+assets bundle (design D3/D4): resolveGentleAiAssetsArchive reads the pinned archive identity from the canonical release lock instead of a second hand-maintained digest table, installAssets stages the signed assets archive into <staging>/assets via the existing bounded release-artifact extractor, and assetsBundleMatches extends the one existing bundle-validity predicate recoverInterruptedPublication already takes as a parameter -- so a binary-without-assets bundle is invalid by construction, with no new publish operation and interrupted-publication recovery covered for free.
…leAiAssets Add the assets provenance keys (assetsAsset, assetsArchiveSha256, assetsTreeSha256, contractMajor, layoutVersion) to the runtime integrity manifest without weakening isCanonicalManifest's exact-key-count/ string-equality discipline, so a missing or forged assets field fails binary resolution exactly like a missing or forged binary field always has. Export resolveGentleAiAssets(packageRoot, platform) for lazy snapshot readers: exact path-set equality against the manifest entries first (an added file a tree digest alone could never catch), then per-file lstat/confinement/mode checks, then a whole-set sameFile TOCTOU recheck across the manifest and every entry file.
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Repository UI Review profile: ASSERTIVE Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Third slice of the consumer chain, stacked on #267. POSIX install path; Windows split provenance is the next slice.
The half-published bundle becomes unobservable, not just unlikely
recoverInterruptedPublicationalready takes the bundle-validity predicate as a parameter and is already called withexistingSignedBundleMatches. Extending that one predicate to cover the assets tree makes a binary-without-assets bundle invalid by construction. No second publish operation, no new crash window.Two tests prove it rather than describe it: an assets-less existing bundle is rejected for reuse and repaired by a fresh full install, and a crash simulated through a fake
renameseam refuses to silently restore a backup missing its assets, failing closed and asking for intervention instead.Path-set equality before digesting, again
Same reason as #266: a digest walk only ever revisits declared entries, so it cannot notice an added file.
resolveGentleAiAssetsasserts exact path-set equality against the manifest first, then per-filelstat, confinement and mode checks, then a whole-setsameFileTOCTOU recheck.The integrity manifest is extended, never loosened
Five new keys join the existing exact key-count and string-equality discipline in
isCanonicalManifest. Every asset file gets its own regular-non-symlink check, not just the binary. Symlink rejection, path confinement and TOCTOU detection are untouched.Threat cases covered: an extra file in the installed tree, a symlinked asset, an asset with mode
0755, an asset namedinstall.shlanding non-executable, TOCTOU replacement mid-verify across the whole file set plus the manifest, and a forged assets digest still rejected byresolveGentleAiBinary.Where the pinned assets digest comes from
The lock, not a second hardcoded table. Reintroducing a hand-maintained table here would recreate exactly the drift that #262 removed for the binary version.
Tests
63/63 focused. Full suite 1083 pass, 3 fail — the known receipt-driven-development-disabled failures, verified identical on clean
mainbefore this chain started.pnpm run check:transaction-runnerclean, runtime regenerated rather than hand-edited.pnpm run test:harnessfails in this sandbox, confirmed by stashing to reproduce identically on the unmodified parent branch. Pre-existing environment condition, same root cause.Rollback
Revert the three source files and two test files to the parent tip and delete the new fixture helper. No file from the earlier or later slices is touched.