Repository navigation
docs(rfc): answer the hybrid RFC's two packaging questions - #22120
Conversation
UX Review (Opus 5.5) — ⏭️ skippedRevision |
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Check that #22065 merges first; both PRs edit the same §7 table, so merging this one out of order will conflict. [DESIGN-REVIEWED] 82b7675 Description read[DESCRIPTION-READ] 856e8d2faa27c2891786a516ae0179b760b3958caba8c8187bb47cfb25d18e14 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
551fab9 to
ffd0d5b
Compare
d5b6c03 to
fc6dcb4
Compare
|
HOLD (base) at fc6dcb4. This PR targets The content itself looks sound: the two §7 rows match the cited code ( — crew-pr-reviewer (r112) |
fc6dcb4 to
918e2c5
Compare
|
Thanks. I've retargeted this to |
Security Scope Review -- 🔴 could not complete (fails closed)No scope verdict for |
|
On the Design note at 918e2c5 (make "confirm once in the first signed build" an actual Phase 3 gate): this PR's frozen scope excludes changing any phase, so I've deferred it rather than widening the PR. For now the gate is the §7 row's Blocks column. @iamwhatever, if you want it in the plan itself, this is the amendment I'd propose for §5's Phase 3 row. Add to the success signal: "in the first signed release, the stub inside the Windows installer passes |
918e2c5 to
57582fd
Compare
Phase 3 of the Python+Rust hybrid RFC was blocked on two questions in section 7: does the Windows signing hook sign an .exe inside a wheel, and does the macOS re-sign script cover a suffix-less Mach-O binary. The code answers both. The desktop installer copies the bundled interpreter's site-packages as an extra resource, and electron-builder hands every .exe there to the sign hook, which exempts only the pinned ffmpeg binary. The released macOS app signs every nested Mach-O by magic. Only the source-install re-sign script misses a suffix-less binary, which Phase 3 now has to extend.
57582fd to
82b7675
Compare
iamwhatever
left a comment
There was a problem hiding this comment.
Same reason as other proposal
Problem / Motivation
Goal: Answer the two packaging questions that block Phase 3 of the Python+Rust hybrid RFC, so the maintainer can settle packaging before any Rust PR moves.
rfc-python-rust-hybrid-backend.md§7 lists two packaging unknowns that block Phase 3'skirocrew-nativewheel. First, does the Windows signing hook sign an.exeinside a wheel ("Unverified"; Windows stays on the Python stub until it is answered)? Second, doespackaging/resign-macos-libs.shcover suffix-less Mach-O binaries (it blocks local macOS installs)? The maintainer asked that packaging be settled before the Rust PRs (#19627, #19628) move.Why it matters
Without these answers Phase 3 would be planned around a Windows exclusion it doesn't need, and it would miss the one change macOS source installs do need.
Not a goal
resign-macos-libs.shchange itself. It belongs with Phase 3, when there is a binary to re-sign.kirocrew-native.What changed (motivation → approach → change)
Lib/site-packagessits inbackend-dist, whichwebsite/electron/package.jsonships asextraResources. In electron-builder 26.15.3,PlatformPackagercopies extra resources throughWinPackager.createTransformerForExtraFiles, andshouldSignFileaccepts every.exeby default.website/electron/scripts/sign-windows.jsthen Authenticode-signs each one, except the pinnedffmpeg-win-x86_64-v7.1.exe. So a stub.exeinside the wheel gets signed with no workflow change. perf(mcp-gateway): Rust MCP stub, used by default when built #19627's embedded source digest survives, because Authenticode appends rather than rewrites; a pinned whole-file hash would not, which is why ffmpeg is exempt.packaging/signing/generate-manifest.pyfinds every Mach-O underContents/Resourcesby magic and signs it with Developer ID, hardened runtime and a timestamp. It already does this forimageio-ffmpeg's suffix-less executable, so the released app is covered.resign-macos-libs.shre-signs only*.soand*.dylib, and onlymake buildruns it. A stub binary would keep its linker-signed signature, so Phase 3 has to extend the script'sfind.rfc-python-rust-hybrid-backend.md§7: both rows now record the answer, its evidence and what still blocks. This branch is stacked on docs(rfc): record measured executor CPU, GIL holders and anon memory #22065, which edits the same table, and should merge after it.Backwards compatibility
Compatible: documentation only.
Removes nothing: the two questions stay in the table with their answers.
Tests
N/A — documentation only.
scripts/docs-lint.shandscripts/check_brand_name.pypass.Manual verification
WinPackager.signApp,createTransformerForExtraFilesandshouldSignFilein the app-builder-lib 26.15.3 package from the public npm registry. Also readPlatformPackager, where extra resources are copied through that transformer.website/electron/package.jsonsets nosignExts, so the default.exerule applies.Related Issues
Stacked on #22065. Context: #19627, #19628.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)Contribution License Agreement
Contributed under the project's Apache-2.0 license.