Repository navigation
Check the installer's download against SHA256SUMS; strip release binaries (MUS-96) - #174
Merged
Merged
Conversation
…ries (MUS-95) install-codex.sh now fetches SHA256SUMS from the same release and checks the shim with sha256sum, or shasum -a 256 on macOS, before it runs it. A mismatch, a missing entry, a missing SHA256SUMS or no tool to check with all stop the install. scripts/test-install-codex.sh runs it against a local fixture through a fake curl, so nothing is downloaded and Codex isn't touched: a matching shim installs, a tampered one is refused and never runs, and the shasum-only path is covered. 9 checks pass; against the installer from before, 7 fail. CI runs it. The release workflow builds with CARGO_PROFILE_RELEASE_STRIP=symbols, so shipped binaries carry no symbols or debuginfo; local release builds are unchanged. Its pins, no-cache rule and smoke test stay. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…be reverted) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…and workflow_dispatch only Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Castor-o7
marked this pull request as ready for review
November 5, 2026 21:42
gilt-skate's open #167 already files MUS-95, so this story is MUS-96, which is unused on main and in every open PR. The README no longer says there's no release pipeline, and says SHA256SUMS comes from the same release, so it catches a corrupt or swapped download, not a compromised release. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Castor-o7
changed the title
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What (MUS-96, follow-ups from MUS-14's review)
(Renumbered from MUS-95, which gilt-skate's #167 already files. The branch keeps its old name.)
scripts/install-codex.shchecks its download. It fetchesSHA256SUMSfrom the same release and checksmuster-shim-<target>withsha256sum, orshasum -a 256where that's all there is (macOS). It refuses to run the shim on any of:SHA256SUMSCARGO_PROFILE_RELEASE_STRIP=symbolsis set on release.yml's build step only, so local release builds keep their symbols. The SHA-pinned actions, the no-cache rule and the smoke test are unchanged.The check catches a corrupt or swapped download, not a tampered release, since both files come from the same release. Signing and notarising wait for Josh's Apple developer setup.
Evidence
Installer:
scripts/test-install-codex.sh(new, run in CI) drives the real installer with a fakecurlserving a local fixture and a "shim" that records how it was run. Nothing is downloaded and Codex is never touched. The checks:codex-setup --daemon …SHA256SUMS→ refusedSHA256SUMS→ refusedshasum: matching installs, tampered is refusedAll 9 checks pass locally. Against main's installer, 7 of the 9 fail.
Stripping: commit 33bcbcc temporarily ran release.yml on this PR, in run 42113899608, and cdce720 reverts it. The tree at cdce720 is identical to 7df2bf5 (
git diff --exit-code). The run passed, and the release job was skipped. Against the unstripped run 42106353650:In the downloaded artifact, all 5 checksums verify,
filereports the Linux binaries as "stripped", and both still run on Murph. Both smoke tests passed on their runners.release.yml changes, so it needs fiery-snapper's read as well as the watchman's.
🤖 Generated with Claude Code