Lagrange v1.0.0: stable (release notes)

v1.0 is the first stable release. Nothing new is simulated: the files, binaries and connections users depend on are frozen for the whole 1.x line, the viewer protocol can be used safely over a network, the packages build as Debian packages and install on a clean machine, and every domain’s validation is written down. Release 1.x is supported until ROS 2 Lyrical’s end of life, May 2031.

Status: released on 2026-10-29 as v1.0.0, under the owner’s standing request to release each version when its work is done. Publishing the packages (ROS buildfarm or an apt repository, the container registry, Sigstore signatures) is the maintainers’ step (packaging/README.md).

Highlights

  • The 1.x freeze (RFC 0007, GOVERNANCE.md). The USD schemas, the plugin C ABI (major 1), the viewer protocol (v1), profiles (schema_version: 1) and recordings (format_version: 1) change only by addition within 1.x. Each surface has a check:

    • schemas: schemas/test/frozen_1_0.json lists all 27 schemas and 240 properties, and the schema test compares against it;

    • plugin ABI: struct layouts are pinned by test_abi_layout.c, and tools/ci/abi_check.sh runs abidiff when libabigail is installed;

    • protocol: buf breaking uses the FILE rules;

    • profiles and recordings: every sample file since v0.5.0 is migrated in a test.

  • Plugin ABI 1.0 adds shift_origin, so a plugin that keeps world positions follows the floating origin.

  • Migrations. migrate_recording joins migrate_schema and migrate_profile. test_old_releases.py checks the sample stages and example profiles of v0.5.0, v0.6.0 and v0.7.0 against this release’s tools.

  • Security (docs/design/security.md).

    • tls:// endpoints carry the viewer protocol over TLS 1.2+ with certificate and host verification.

    • A token (--auth-token-file) is compared in constant time, and is required to listen beyond loopback.

    • Unix sockets stay mode 0600.

    • The scene builder has a fuzz harness (libFuzzer, plus a mutation test of 54 000 frames in every build).

    • The threat model and dependency review are in security.md.

  • Binary packages, built locally (packaging/README.md).

    • Every package as a ros-lyrical-lagrange-* .deb, built by bloom in a clean container.

    • A release image from the debs alone, which is also the clean-machine install test.

    • A Python wheel and an SPDX 2.3 SBOM.

  • Validation report (docs/validation/report.md): per domain, what each model is checked against, the tolerance, the result and the test, and where sim-vs-real data is still owed.

Assets

  • lagrange-sedan-city.mp4: 20 s of the whole viewer with the sample sedan in the city block (the README’s clip, attached to the GitHub release because the repository keeps files under 1 MB).

  • The debs, the release image, the wheel and the SBOM are built with packaging/ and not attached: publishing them is the maintainers’ step.

Acceptance (roadmap §9)

Epic

Acceptance

Status

Evidence

Freeze

Freeze RFC accepted

met

RFC 0007, accepted under the owner’s 2026-10-29 pre-approval; the checks above. buf and libabigail are not installed on the development machine, so buf breaking and abidiff did not run; the layout test and the frozen schema list run in every build

Migration tools

Old sample files migrate in CI

met, locally

test_old_releases.py over v0.5.0–v0.7.0 (GitHub Actions are disabled)

Binary releases

Clean-machine install test

met, locally

All 46 packages built as .debs by bloom in a clean ros:lyrical-ros-base-resolute container; packaging/test_install.sh installed them alone into a fresh image and ran the CLI, a headless run from Python, the determinism check, a streamed world, the migration tools and the world builder (Tier 0, llvmpipe). Built from the release candidate before the version bump, so those debs say 0.7.0; rebuild from the v1.0.0 tag for publishing

Documentation

All tutorials run in CI

met, locally

tools/ci/tutorial_{amr,drone,environment,cars,huge_worlds}.sh and arm_scenario.sh

Validation report

Published with the release

met

docs/validation/report.md. Sim-vs-real against measured data exists for the battery only; the rest waits for open question Q16

LTS policy

Documented in GOVERNANCE.md

met

release/1.x until May 2031, backport rules

Security review

Review issues closed

met, with open items

docs/design/security.md §5: an advisory scan of the vendored packages and long fuzzing campaigns are owed

Known issues

  • Not run while GitHub Actions are disabled: the sanitizer builds, the nightly jobs, buf breaking, abidiff, long fuzzing campaigns.

  • Owed from earlier releases: everything listed as owed in docs/plan/progress.md and in the validation report (sim-vs-real data, ISO manoeuvres, ASAM scenarios, renderer scale for huge worlds, moving robots between regions while running).

  • Radiation stays a zero substitute (post-1.0, the owner’s decision).

Release checklist

  • [x] Every v1.0 epic done or with its open items listed

  • [x] Release notes

  • [x] Gated changes listed for the owner’s review (progress.md: RFC 0007, the frozen schemas and ABI, ClientHello.auth_token)

  • [ ] Owner review of those merges

  • [ ] GitHub Actions re-enabled and green on dev

  • [ ] Packages published and signed (maintainers, packaging/README.md)

  • [x] Released: dev merged into main, tagged v1.0.0, and release/1.x branched