zeekay
04ce8745ee
Squashed batch covering all post-Feb-28 consensus work, capped by the LSS-Pulsar cutover (task #53): - Quasar cutover from upstream Corona to LSS-Pulsar adapter: go.mod: drop direct require github.com/luxfi/corona; add github.com/luxfi/pulsar + github.com/luxfi/threshold; local replaces => ../pulsar, ../threshold epoch.go: InitializeEpoch → keyera.Bootstrap; RotateEpoch → lss.DynamicResharePulsar gated by activation cert verification under unchanged GroupKey quasar.go, grouped_threshold.go: import swap corona → pulsar reshare_epoch.go: production wiring; QUASAR-PULSAR-ACTIVATE-v1 domain prefix; ReshareEpoch + RefreshEpoch EpochKeys: gains KeyEraID, Generation, RollbackFrom (LP-105) BundleSigner: re-derives Lambda for participating subset 258 quasar tests pass. - Earlier post-Feb-28 work (gofmt linter fixes across protocol files, METASTABLE_SIGNED_QUERIES.md doc, MLDSAProof rename across 7 test files) included in the squash. Hash profile: consensus passes nil to resolve to Pulsar-SHA3 default (LP-105 / pulsar/hash). Activation message canonical bytes bind chain_id, network_id, key_era_id, generation, group_id, epoch, set hashes, thresholds, group_pk_hash, nebula_root, hash_suite_id, implementation_version, variant, reshare_transcript_hash — all under TupleHash256("PULSAR-TRANSCRIPT-v1", ...). Vocabulary alignment with LP-105 (Lux Finality Stack lexicon): Photon (vote) → Beam (BLS) → ML-DSA (PQ identity) → Pulsar (PQ threshold kernel) → Pulse (cert over Nebula root) → Prism (composition verifier) → Horizon (finality boundary) → Quasar (full consensus). Excluded from this commit: pkg/wire/{candidate,policies_test}.go uncommitted diff (pre-existing local edits; stashed and restored). Untracked protocol/quasar/witness_producer.go (mtime Apr 29; pre-task).
4.3 KiB
Cross-Repo Version Pin — March 3, 2026 PQ Consensus Architecture Freeze
This document is the canonical record of which commit SHAs the
v0.1.0-rc1-pq-consensus-freeze tag points at in every repo that
participates in the Lux post-quantum consensus stack.
Tag
Every repo below is tagged annotated with:
git tag -a v0.1.0-rc1-pq-consensus-freeze \
-m "Mar 3 2026 Lux PQ Consensus Architecture Freeze (locked scope)"
Repos and SHAs
| Repo | Path | Tag | Commit SHA |
|---|---|---|---|
| pulsar | ~/work/lux/pulsar |
v0.1.0-rc1-pq-consensus-freeze |
7b3a53f0ba569a7092fe969c759bfbb28f5341ec |
| lens | ~/work/lux/lens |
v0.1.0-rc1-pq-consensus-freeze |
72d0dd6c470caab0689ca1a1e59881c044cc5cb6 |
| threshold | ~/work/lux/threshold |
v0.1.0-rc1-pq-consensus-freeze |
e4e6be547b7db23726f0b18474bd00ea9801592c |
| consensus | ~/work/lux/consensus |
v0.1.0-rc1-pq-consensus-freeze |
853c48cb3149fd93b9517a7818a6200b4d9557f4 |
| warp | ~/work/lux/warp |
v0.1.0-rc1-pq-consensus-freeze |
d4237f861b6b28aca6a6ccb8693a65a769205b18 |
| proofs | ~/work/lux/proofs |
v0.1.0-rc1-pq-consensus-freeze |
cf682ab2fb1b233331f0880d6496f05883ff6583 |
| papers | ~/work/lux/papers |
v0.1.0-rc1-pq-consensus-freeze |
3529523e60c0f03550705cdd526bd2258db71c58 |
| lps | ~/work/lux/lps |
v0.1.0-rc1-pq-consensus-freeze |
c582b6d3e3c2dc0e3904d211a13c6b6e0a962c42 |
These SHAs are produced by amending each repo's existing 2026-03-03 HEAD with the gate-1/2/5 deliverables and tagging that amended commit. The annotated tag message is: "Mar 3 2026 Lux PQ Consensus Architecture Freeze (locked scope)".
The consensus row records the SHA of the amend that immediately
precedes the final SHA-table commit (since this file LIVES IN that
final commit and would otherwise need to record its own SHA, which
is a chicken-and-egg). Run git log -1 --format=%H v0.1.0-rc1-pq-consensus-freeze
inside ~/work/lux/consensus to see the actual tagged commit.
go.mod Pinning
The pulled-in repos (consensus, threshold, warp) declare the tag as
their require version and keep the local-dev replace directive
pointing at the sister checkout. Replace lines carry a comment:
// pinned to v0.1.0-rc1-pq-consensus-freeze; see go-mod-pin.md
This lets local development proceed against the working tree without relying on a published tagged module proxy. CI builds without the neighbouring checkouts use the tagged version directly.
consensus/go.mod
require (
github.com/luxfi/pulsar v0.1.0-rc1-pq-consensus-freeze
github.com/luxfi/threshold v0.1.0-rc1-pq-consensus-freeze
github.com/luxfi/lens v0.1.0-rc1-pq-consensus-freeze // indirect
)
replace github.com/luxfi/pulsar => ../pulsar // pinned to v0.1.0-rc1-pq-consensus-freeze; see go-mod-pin.md
replace github.com/luxfi/threshold => ../threshold // pinned to v0.1.0-rc1-pq-consensus-freeze; see go-mod-pin.md
replace github.com/luxfi/lens => ../lens // pinned to v0.1.0-rc1-pq-consensus-freeze; see go-mod-pin.md
threshold/go.mod
require (
github.com/luxfi/lens v0.1.0-rc1-pq-consensus-freeze
github.com/luxfi/pulsar v0.1.0-rc1-pq-consensus-freeze
)
replace (
github.com/luxfi/lens => ../lens // pinned to v0.1.0-rc1-pq-consensus-freeze; see go-mod-pin.md
github.com/luxfi/pulsar => ../pulsar // pinned to v0.1.0-rc1-pq-consensus-freeze; see go-mod-pin.md
)
warp/go.mod
require github.com/luxfi/pulsar v0.1.0-rc1-pq-consensus-freeze
replace github.com/luxfi/pulsar => ../pulsar // pinned to v0.1.0-rc1-pq-consensus-freeze; see go-mod-pin.md
Verifying the Pin
A canonical proof point is whether ~/work/lux/scripts/regen-all-kats.sh --verify passes against the tagged commit on each repo. That script
runs:
pulsar/scripts/regen-kats.sh --verifylens/scripts/regen-kats.sh --verifywarp/scripts/regen-kats.sh --verifythreshold/scripts/regen-kats.sh --verify
and asserts byte-equality of all KAT outputs against the manifest each repo's regen-kats.sh writes.
Forward Path
The next freeze should bump the tag (e.g. v0.1.0-rc2-...) and
update both the require lines and the comments on the replace
lines. The replace directives stay pointing at the local working
tree — they're the local-dev affordance and never go away.