feat(cast-render): build v9 for arm64 and verify each architecture #361
Loading…
Reference in a new issue
No description provided.
Delete branch "feat/cast-render-arm64"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Why
The fleet has only ARM64 workers since 2026-09-12. The ibmi-rpg recipe in
binjovi-planspinsFROM --platform=linux/amd64 registry.sean.farm/build/cast-render:v8@sha256:fb1fb122...and renders the reels under an emulator. Today a release of rpg-munchers failed in that stage after 219 s with.buildkit_qemu_emulator: Invalid ELF image for this architecture.What
cast-renderv9. Binjovi builds it forlinux/amd64andlinux/arm64from the one Dockerfile (the agent's dockerfile-set allowlist passesplatform=linux/amd64,linux/arm64for this image). The plans recipe can move to v9 and drop its--platformpin.TARGETARCHagg asset selection (agg-x86_64-unknown-linux-gnu,agg-aarch64-unknown-linux-gnu). Cite the release page and the asset digests for both sha256 pins. Both assets were downloaded again on 2026-09-14 and checked withshasum -a 256.glibcandffmpeg-freemust have theTARGETARCHpackage architecture, no foreign-architecture package may be installed, and the ELF machine field ofagg,ffmpegandffprobemust matchTARGETARCH. An emulator runs a wrong binary without an error; this gate does not.aggand encode it withlibopenh264in the gate. This proves the monospace font and the real H.264 encoder with the same commands the ibmi-rpgreels-renderstage runs.mc. No consumer runsmcinside this image: the ibmi-rpg recipe and the historical import publish media from the package-publisher image.--proto '=https'and retries. Point the source label at code.sean.farm.The
TARGETARCHversusuname -mcheck, the ffmpeg/ffprobe H.264 checks, and thePKG_CACHEflow are unchanged.Checks
bash scripts/lint-no-runtime-fetch.sh dockerfiles images: clean.agg(ELF machine 0xb7, read from the published layer). v9 makes that a build-time gate.https://claude.ai/code/session_01HPdKaFBKQVs8hzfgXMrnwV