* windows: sign the setup engine so smart app control can run setup
Setup extracts a copy of its engine to %TEMP% at launch and writes the
installed uninstaller from the same bytes. Signing only the outer
hex-setup-*.exe left both unsigned, and Windows 11 Smart App Control
blocked the extracted copy ("Unable to execute file in the temporary
directory", error 4551).
SignedUninstaller embeds the engine's signature, but the signature has to
exist before the compile that uses it, and the compile job holds no
credentials. So the release signs it in two passes: the new
windows-setup-engine.yml compiles once with SignedUninstaller=yes and keeps
the unsigned uninst-*.e32 ISCC asks to be signed (ISCC leaves it and aborts
as a compile failure, exit 2); sign-windows-engine signs and verifies it in
the release environment; windows-installers.yml then compiles with the
signed engine and asserts the installed uninstaller's signature.
Test builds and runs without the Azure values keep the unsigned single
compile. Inno Setup is pinned to 6.7.1, read from the compiler's own banner
(ISCC.exe's version resource is 0.0.0), so both passes use one compiler.
* windows: harden the setup engine passes
Review fixes for the signed setup engine:
- sign-windows-engine requires sign-windows to have succeeded, and its
steps are unguarded like sign-windows-installers, so it can never upload
an empty digest.
- setup-engine runs only when Windows is signed (it waits for
sign-windows), so unsigned runs really do a single compile with no
engine pass at all.
- the Inno re-pin uses choco upgrade --allow-downgrade, because choco
install --version treats a preinstalled package as already installed.
- Invoke-Iscc collects native stderr safely under Windows PowerShell 5.1
and pwsh with $PSNativeCommandUseErrorActionPreference on, and echoes
the compiler's lines so build logs stay readable.
- the version probe compiles a minimal script from standard input into a
throwaway OutputDir, so it never touches the caller's -OutDir.
- only .e32 exists for the engine; drop the unused .e64 handling.
- the uninstaller signature test also checks the release publisher, and
the test-builds doc warns that Smart App Control cannot be turned back
on without resetting Windows.
* windows: correct the setup engine comments
KDE's color-scheme portal describes window colors, so a mixed theme such
as Breeze Twilight picked a black mark for a dark panel. Bump rune for
TraySpec::icon_name, write anomaly.hex-symbolic.svg into the user's
hicolor theme at startup, and serve that name beside the pixmap so Plasma
6, GNOME's AppIndicator extension, and Waybar draw the mark in their own
text color. The icon keeps the old white/black default for hosts that
don't recolor, and the installer removes only a marked file.
Goldens follow the rune bump.
The site deploy job ran in the release environment and assumed role/github/hex-release, whose token could write every Worker on the account. It now runs in infra-dev or infra-production (picked from the mode, each admitting only its own branch) and assumes role/github/hex-infra-<mode>, which mints a token for that mode's Worker only. Release jobs stay in release.
The deploy job assumed role/github/hex-release and then read the long-lived
hex/release/CLOUDFLARE_API_TOKEN from Secrets Manager. It now mints a
15-minute token from the token issuer after assuming the role, masks it
before writing $GITHUB_OUTPUT, and passes it only to the Deploy step.
* ci: read release secrets from aws secrets manager
* ci: read the linux signing key from hex/release/RELEASE_SIGNING_KEY
* ci: hex's own aur and homebrew keys; the signing cert has no password
* ci: the signing certificate has no password
* ci: use the shared aur and homebrew keys
* docs: the shared homebrew deploy key is anomaly-release
* docs: the shared homebrew deploy key is anomaly-release
* site: own /hex/files, drop the r2 write credential (#1611)
* site: own /hex/files for installers, appcast, and dev notes
The site build now copies install-linux.sh, install-windows.ps1, and
packaging/macos/legacy-appcast.xml into its Worker assets, so the same
Worker serves them at /hex/files with the Content-Type and Cache-Control
the bucket objects had. The Installers and Legacy appcast workflows,
r2-put.sh, and their R2 write credential are gone.
The dev notes move to the one-commit dev-notes branch, published by
Release's dev-notes job with GITHUB_TOKEN only; the Worker proxies that
file at /hex/files/dev-notes.json (falling back to the bucket copy until
the first dev release creates the branch) and renders /hex/dev from it.
Both site Workers now bind hex-production-files, read-only in code, for
the frozen 2.3.1 feeds and artifacts, which lets hex-dev-files go. The
Site build runs the installer test and appcast check the removed
workflows ran, and the verify job compares the served bytes.
* docs: the site owns /hex/files and needs no R2 write credential
Hosting, secrets, old clients, and Linux packages describe the built
installers and appcast, the dev-notes branch, and the read-only bucket.
* fix(linux): name the missing wl-copy before a Wayland paste
* fix(linux): address review on wl-copy warning
* fix(settings): show the wl-copy notice apart from dictation failures
* Windows: ship hex.com, a console entry point beside the GUI hex.exe, on PATH
* lock, fmt, legacy adoption
* clippy
* build hex-console in its own cargo run: -p hex is ambiguous with the hex crate
* hex.com returns once the app starts; Ctrl+C only; install script and Setup replace a running hex.com; hex-console checks in Compile
* fmt
* installer test: exit 0 once every check passed
* windows: delay-load vulkan-1.dll and fall back to CPU without a Vulkan 1.2 driver (#1240)
* windows: note vulkan-1.dll is delay-loaded in restrict_dll_search
* Linux: use the Background portal for running in the background and Launch at login (#812); fallback entry written once (#810)
* test names
* Docs: Background portal and Launch at login (#812)
* Background portal: only once registered (1.20+), symlinked entries are the user's, serialize entry writes and login changes
* doc comment
* Dev builds say what's new: notes from PR titles in the dev feed, /hex/dev, and Restart to update (#763)
* Dev notes: keep ReadyToRestart's wire format; canonical upload; script and refetch fixes from review
* Sidebar update in its own module (code-shape)
* x86_64: keep ggml on x86-64-v3 and refuse older CPUs with a clear message (#592)
Reverts #595's Ivy Bridge floor: it slowed CPU transcription for everyone. ggml goes back to its x86-64-v3 defaults (AVX2, BMI2, FMA, F16C). Hex's own Rust stays baseline x86-64, and the first thing main does on Linux and Windows x86_64 is check AVX2/BMI2/FMA/F16C with is_x86_feature_detected!. Without them Hex prints and shows "Hex needs a 2013 or newer processor (Intel Haswell / AMD Excavator or later) with AVX2. This PC's processor isn't supported." (a native MessageBox on Windows; stderr, which the journal logs, and zenity, kdialog or notify-send on Linux) and exits 1. Every command is covered except --version and --help, which installers and the updater run. ggml-cpu has no static constructors, so nothing runs v3 code before main.
The release guard now asserts the opposite: ggml's CPU libraries must hold AVX2 and FMA (the build targets v3) and no AVX-512 or AMX. Docs state the requirement.
* cpu_floor: harden DLL search first; no dialog under systemd, in a terminal or for hex service; guard rejects AVX-VNNI
* cpu_floor: always exit 1; Windows always shows the message box; Linux skips the dialog in a terminal, under NOTIFY_SOCKET, or for hex service
* Lowercase release file names; README downloads by channel (#456)
Release builds and the GitHub release mirror now name files hex-<version>-macos-arm64.{zip,dmg}, hex-<version>-windows-<arch>.exe and hex-setup-<version>-windows-<arch>.exe (Linux was already lowercase), as @vixalien asked in #456. Already-published files and the frozen 2.3.1 feeds keep their names; the release service's published names (hex, Hex.dmg, Hex.zip, hex.exe, HexSetup.exe) and the bundle (Hex.app) are unchanged, so the updater, installers, Homebrew cask, AUR and Nix are unaffected.
The README downloads table has Stable and Dev columns for every target, including the updater files. The site Worker serves /hex/download/dev/<target> (the dev alias) beside /hex/download/<target> (stable, or dev while there is none), and both now cover the bare updater files.
* Lowercase names in two remaining examples
* Lowercase the Windows executable in the CPU check step
* x86_64 releases: build ggml for the Ivy Bridge floor and fail on BMI2/AVX2/FMA (issue #592)
GGML_NATIVE=OFF alone leaves ggml's GGML_AVX2/BMI2/FMA/F16C on, so Hex 3.0.0 on Windows x64 died with 0xc000001d (illegal instruction) on an Ivy Bridge Xeon E3-1220 V2 once transcription started (reported by @Hc-Web-Dev). Linux and Windows x86_64 release builds now pin SSE4.2 + AVX + F16C and turn AVX2, BMI2, FMA, AVX-VNNI, AVX-512 and AMX off (scripts/ggml-x86-baseline.txt), the floor transcribe.cpp already uses on Intel Macs. Runtime dispatch (GGML_BACKEND_DL + GGML_CPU_ALL_VARIANTS) needs shared ggml libraries, which our single static executable can't ship. CPU transcription on newer x86 CPUs loses AVX2/FMA kernels; Vulkan is unaffected.
.github/release/check-x86-baseline.sh guards both build jobs: the executable may hold no BMI2 instruction (3.0.0's Linux build holds 100; Rust crates that use AVX2/FMA/AVX-512 behind run-time detection use none), and the ggml static libraries no AVX2, BMI2, FMA or AVX-512. No crate sets target-cpu; the only rustflags are +crt-static; no other C/C++ dependency builds with -march=native or AVX2 (Moonshine/ONNX are gone).
* check-x86-baseline: fail on missing inputs or empty disassembly; match llvm-objdump's suffixed BMI2 mnemonics
#424 removed Moonshine (and #430 Smart Turn's onnxruntime), but sign-macos.sh still required libmoonshine.dylib, so every Release since 0.0.0-dev-249 failed at sign-macos and nothing published. The bundle's code allowlist now requires only hex, scrawl-helper, and the microphone driver, and accepts libonnxruntime only if a build still bundles it. On the dev channel, publish no longer waits on macOS: when signing or verification fails, Linux and Windows still publish (only a verified macOS build is ever included), and the run still fails visibly. Stable still requires every platform.
From the 2026-10-29 security review:
- H1: Release signs each Linux executable with Anomaly's ed25519 key (HEX_RELEASE_SIGNING_KEY, .github/release/sign-linux.sh, which refuses a key that does not match linux-signing.pub) and publishes the signature as the linux-<arch>-sig target. The updater (updater/linux.rs SIGNING_KEY) and install-linux.sh refuse an executable without a valid signature over its SHA-256, before downloading it. Installed builds that predate this ignore the extra target, so the transition needs nothing from them.
- H3: the installers move to the locked .github/release/ (CODEOWNERS @thdxr); the Installers workflow publishes them from a sparse checkout of that directory.
- M1: the release service client is vendored as .github/release/release-client.sh instead of piping anoma.ly/release/client.sh into bash.
- M2: check-windows.yml gets permissions: contents: read; every third-party action is pinned by SHA.
- M3: resolve_url accepts only URLs on the manifest's host.
build-macos ships the background TIFF and Finder's .DS_Store beside Hex.app.zip; sign-macos.sh checks both by magic number and size and copies them into the DMG as data.
curl's own --retry reused the request's token, so a retry after a slow or
stalled upload could still hit 401 token expired. Retries now happen in the
script (up to 4 attempts on a failed transfer or a 429/5xx, each logged),
every attempt mints its own token, and a transfer stalled below 1 KB/s for a
minute is abandoned and retried. Multipart part PUTs and the finalize POST go
through the same path. Ports anomalyco/hokum 86d5ce3.
Run 40785730883's publish failed with 401 token expired: release-service.sh minted one GitHub OIDC token (they expire within minutes, and the service enforces exp) and used it for every upload, and the uploads took 7.5 minutes (hex.deb 2.5 min, hex.rpm 4 min). Each request now mints its own token, uploads run in parallel, a file the service already stores under the same sha256 is skipped, and each upload logs its duration.
* Release: build hex-bin in builder's home instead of chowning the runner's files
The AUR job chowned RUNNER_TEMP to builder, taking the runner's own files
(step summary, event.json) with it, so the summary, upload and post-checkout
would fail with permission denied on the first stable release. builder now
stages and runs makepkg on copies in /home/builder; the workspace and
RUNNER_TEMP keep their owner.
* Release: upload hidden files so build-aur keeps its .SRCINFO
* Linux ARM64, .deb, and .rpm releases
Both Linux architectures build natively on Ubuntu 24.04 runners (glibc 2.39
floor, Vulkan on both), publish linux-arm64 plus linux-<arch>-deb and -rpm
(hex.deb, hex.rpm) built with nfpm and installed on Ubuntu and Fedora in the
build job. hex-bin gains aarch64. Package installs report deb-<arch> or
rpm-<arch> through dpkg/rpm ownership and get the stable update command.
* build-linux.sh prints only the executable's path
* Linux releases link no system BLAS, as the Arch build shipped
* Linux build: keep libopenblas-dev for build.rs's dropped -lopenblas
Security finding #528: Site ran site/ code (bun install, build.ts, vite)
in the release environment with a Cloudflare token that can write every
Worker. Site deploy now builds without an environment or secret, uploads
the Build Output with its SHA256SUMS digest, and deploys it from a job
that checks out only .github/release/, verifies the digest, and runs
deploy-site.sh: it refuses any Worker, route, or binding other than the
mode's site Worker and deploys with the cf CLI locked in
.github/release/cf/. Legacy appcast checks out only .github/release/ and
the appcast, and r2-put.sh moves under the boundary.
The deploy token is now hex-site-deploy (Scripts + anoma.ly Routes, no
R2) and the appcast key hex-r2-production-files (one bucket).
release.yml restarted GitHub's run numbers at 1, so the first run from it published 0.0.0-dev-1 and later runs would have reused 0.0.0-dev-32…45. macOS identifies a build by its version and Linux/Windows name version directories after it, so a reused version could leave a Mac on an old build or collide with a running version on Windows. Dev versions are now 0.0.0-dev-<run number + 100>.
The release service stores each file at files/<sha256>/<name> and serves it as an attachment named <name>, so names need no version or target: linux-x64 is hex, macos-arm64-dmg Hex.dmg, macos-arm64 Hex.zip, windows-x64/arm64 hex.exe, and the Windows installers HexSetup.exe (the service's names allow no spaces). Targets reference {file, sha256}, so the two Windows architectures may share a name. Builds and the GitHub release mirror keep versioned names; release-service.sh maps them. The updater, installers, and AUR PKGBUILD (hex-<pkgver>-linux-x64::<url>) take the URL from the manifest; the AUR and Homebrew renderers accept the new URLs, and the cask URL no longer interpolates the version.
Tags trigger nothing, so v* tags are unrestricted. The release deploy key, its HEX_RELEASE_TAG_KEY secret, and the tag ruleset are gone. The final job (contents: write, no environment) pushes v<version> on the released commit with the built-in token after everything published, then creates the GitHub release; a check run pushes and deletes a throwaway v0.0.0-test-<run> tag with the same token.
Release no longer runs on tag pushes; stable is a dispatch from production only and refuses a version already tagged v<version> or app-v<version>. After every platform and the release service published, a tag job pushes v<version> on the released commit with the release deploy key (HEX_RELEASE_TAG_KEY, admitted by the new v* tag ruleset), then the GitHub release mirrors the files on that tag; a failed run leaves no tag. A check run pushes and deletes a throwaway v0.0.0-test-<run> tag to prove the ruleset admits the key. The release environment now admits only dev and production.
download-artifact unpacks a single matching artifact straight into the download folder; verify-artifacts.sh now files it under the name whose reported digest its SHA256SUMS has. The Windows build checks out into source/ because the Vulkan SDK installs into the workspace. verify-macos runs only after sign-macos succeeds. .github/CODEOWNERS names @thdxr; docs describe the split release.
Build jobs (Linux, Windows x64/ARM64, macOS, Windows installers, AUR and Homebrew staging) run repository code with no environment, secret, or id-token and produce unsigned files. Sign and publish jobs run in the release environment, check out only .github/release/, and run only those scripts and SHA-pinned actions on artifacts treated as data: macOS Developer ID signing, notarization, and stapling (from APPLE_SIGNING_CERTIFICATE, APPLE_NOTARY_KEY, and the APPLE_NOTARY_KEY_ID/ISSUER_ID variables), Azure Authenticode, the release service, the GitHub release, and the AUR and Homebrew pushes. Every artifact carries a SHA256SUMS whose digest its job reports; every job that downloads it checks that digest. Signed macOS builds are verified and launched in a job without secrets. macos-signed.yml is gone (a check run covers it); the site deploy is no longer part of Release.