-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathDockerfile
More file actions
740 lines (704 loc) · 46.3 KB
/
Copy pathDockerfile
File metadata and controls
740 lines (704 loc) · 46.3 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
# The version is supplied as a build argument rather than hard-coded
# to minimize the cost of version changes. Must be >= the `go` directive
# in go.mod (1.26.4); the EVM plugin pulls luxfi/upgrade@v1.0.1 which
# floors the toolchain at 1.26.4. Pinned to the latest stable patch so the
# image ships current compiler/stdlib security fixes; the builder stage also
# sets GOTOOLCHAIN=auto so a future floor bump downloads rather than fails.
ARG GO_VERSION=1.26.5
# ============= Go Installation Stage ================
FROM --platform=$BUILDPLATFORM debian:bookworm-slim AS go-installer
RUN apt-get update && apt-get install -y --no-install-recommends \
wget ca-certificates \
&& rm -rf /var/lib/apt/lists/*
ARG GO_VERSION
ARG BUILDPLATFORM
# Download Go for build platform
RUN BUILDARCH=$(echo ${BUILDPLATFORM} | cut -d / -f2) && \
wget -q "https://go.dev/dl/go${GO_VERSION}.linux-${BUILDARCH}.tar.gz" && \
tar -C /usr/local -xzf "go${GO_VERSION}.linux-${BUILDARCH}.tar.gz" && \
rm "go${GO_VERSION}.linux-${BUILDARCH}.tar.gz"
# ============= Compilation Stage ================
# Always use the native platform to ensure fast builds
FROM --platform=$BUILDPLATFORM debian:bookworm-slim AS builder
# Copy Go from installer stage
COPY --from=go-installer /usr/local/go /usr/local/go
ENV PATH="/usr/local/go/bin:${PATH}"
# A go.mod `go` directive newer than the toolchain above is a hard failure
# under GOTOOLCHAIN=local. `auto` lets go fetch the required toolchain
# (checksum-verified via sum.golang.org) instead of dying at the floor.
ENV GOTOOLCHAIN=auto
# Install build dependencies (ca-certificates needed for go mod download)
# libc6-dev-arm64-cross needed for cross-compiling to ARM64
RUN apt-get update && apt-get install -y --no-install-recommends \
gcc libc6-dev make git ca-certificates wget \
gcc-aarch64-linux-gnu gcc-x86-64-linux-gnu \
libc6-dev-arm64-cross libc6-dev-amd64-cross \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /build
# Skip checksum verification for luxfi + hanzoai packages: both are cross-org
# deps not registered in the public sum.golang.org / proxy (reading e.g.
# hanzoai/vfs@v0.4.1's go.mod via the public sumdb 404s and fails the build).
# GOPRIVATE is the one that governs checksum verification; GONOSUMCHECK predates
# modules and Go ignores it. Our own modules must be listed here: a tag that was
# ever moved no longer matches what the public checksum database recorded for it,
# and every later build that resolves that version dies on
# verifying go.mod: checksum mismatch / SECURITY ERROR
# even though nothing in the build is wrong. The plugin stages resolve transitive
# luxfi versions we do not choose, so the exclusion has to cover the whole org.
ENV GOPRIVATE=github.com/luxfi/*,github.com/lux-private/*,github.com/hanzoai/*
# GONOPROXY is narrower than GOPRIVATE on purpose: keep fetching through the proxy
# (gonum.org is flaky direct), and go direct only for the modules the proxy cannot
# see. Setting GOPRIVATE alone would silently take everything off the proxy.
ENV GOPROXY=https://proxy.golang.org,direct
ENV GONOPROXY=github.com/lux-private/*,github.com/hanzoai/*
ENV GOFLAGS="-mod=mod"
# Copy and download lux dependencies using go mod.
# Some luxfi/* modules (e.g. corona) are private and require a token to
# resolve via go mod. The token is injected as a BuildKit secret from the
# CI workflow; locally, set DOCKER_BUILDKIT=1 and pass
# `--secret id=ghtok,src=$HOME/.gh-token`. If the secret is absent the
# build still attempts the download (works when all deps are public).
COPY go.mod .
COPY go.sum .
# Every module this builds from is PUBLIC — go.mod names nothing under
# lux-private/ or luxcpp/ — so the module download authenticates to nothing and
# fetches anonymously. The credential below is scoped to the two private orgs
# only, so that a private module added later resolves without this step having
# to be rediscovered, while github.com at large stays anonymous.
#
# Scoping is the point, not tidiness. Three release builds failed here:
#
# * Twice because the token was interpolated into a URL's authority —
# `url."https://x-access-token:$(cat …)@github.com/".insteadOf` — where a
# single `/` in the token ends the authority early and git resolves the
# credential prefix as a hostname:
# Could not resolve host: x-access-token
# which reads as DNS rather than as quoting.
#
# * Once because the fix for that sent the token as an Authorization header
# for ALL of github.com. The token is a registry credential: it logs in to
# ghcr.io, and it cannot read repositories. Presenting it made GitHub answer
# 401 for github.com/hanzoai/vfs — a PUBLIC repository that had been served
# anonymously all along:
# fatal: could not read Username for 'https://github.com'
# A wrong credential is worse than none, because none is a 200.
#
# base64 is used rather than URL interpolation because a header value has no
# character that can truncate it, so this stays correct when the token is next
# rotated to a different shape.
RUN --mount=type=secret,id=ghtok,required=false \
if [ -s /run/secrets/ghtok ]; then \
auth="Authorization: Basic $(printf 'x-access-token:%s' "$(cat /run/secrets/ghtok)" | base64 | tr -d '\n')"; \
git config --global http."https://github.com/lux-private/".extraheader "$auth"; \
git config --global http."https://github.com/luxcpp/".extraheader "$auth"; \
fi && \
sed -i -E '/^github.com\/(luxfi|hanzoai)\//d' go.sum && \
go mod download -x
# Copy the code into the container
COPY . .
# Ensure pre-existing builds are not available for inclusion in the final image
RUN [ -d ./build ] && rm -rf ./build/* || true
ARG TARGETPLATFORM
ARG BUILDPLATFORM
# Per SCALE_STANDARD.md §2 (https://github.com/hanzoai/hips/blob/main/docs/SCALE_STANDARD.md)
# — every Go production Dockerfile that emits JSON to a client builds
# with GOEXPERIMENT=jsonv2. Verified -12% time / -23% allocs on the
# edge POST roundtrip vs encoding/json v1. Applies to luxd + every
# in-stage VM plugin build below.
ARG GO_EXPERIMENT=jsonv2
ENV GOEXPERIMENT=${GO_EXPERIMENT}
# Configure a cross-compiler if the target platform differs from the build platform.
#
# build_env.sh is used to capture the environmental changes required by the build step since RUN
# environment state is not otherwise persistent.
RUN if [ "$TARGETPLATFORM" = "linux/arm64" ] && [ "$BUILDPLATFORM" != "linux/arm64" ]; then \
echo "export CC=aarch64-linux-gnu-gcc" > ./build_env.sh \
; elif [ "$TARGETPLATFORM" = "linux/amd64" ] && [ "$BUILDPLATFORM" != "linux/amd64" ]; then \
echo "export CC=x86_64-linux-gnu-gcc" > ./build_env.sh \
; else \
echo "export CC=gcc" > ./build_env.sh \
; fi
# Fetch pre-built lux-accel (GPU crypto library). The release assets live in
# the PRIVATE luxcpp/accel repo (resolves to lux-private/accel) — an
# unauthenticated GitHub release-download 404s, so the fetch is authenticated
# with the same `ghtok` BuildKit secret used for private go modules. It is
# also best-effort (matches the cevm/lpm fetch contract below): the library is
# ONLY linked at CGO_ENABLED=1, so a CGO_ENABLED=0 build (the canonical CI/devnet
# build, pure-Go) does not need it and must not fail when it is unreachable.
ARG ACCEL_VERSION=v0.1.0
RUN --mount=type=secret,id=ghtok,required=false \
ARCH=$(echo ${TARGETPLATFORM} | cut -d / -f2) && \
if [ "$ARCH" = "amd64" ]; then ACCEL_ARCH="linux-x86_64"; else ACCEL_ARCH="linux-arm64"; fi && \
mkdir -p /usr/local/include /usr/local/lib && \
AUTH=""; [ -s /run/secrets/ghtok ] && AUTH="--header=Authorization: Bearer $(cat /run/secrets/ghtok)"; \
( wget -q ${AUTH:+"$AUTH"} \
"https://github.com/luxcpp/accel/releases/download/${ACCEL_VERSION}/lux-accel-${ACCEL_ARCH}.tar.gz" \
-O /tmp/accel.tar.gz \
&& tar -xzf /tmp/accel.tar.gz -C /usr/local \
&& rm /tmp/accel.tar.gz \
&& ldconfig 2>/dev/null \
) || echo "WARN: lux-accel ${ACCEL_VERSION} fetch skipped (private/unreachable; GPU accel unused at CGO_ENABLED=0)"
# Fetch pre-built luxcpp/cevm libs (libevm, libevm-gpu, libluxgpu,
# libcevm_precompiles + go_bridge.h). When CGO_ENABLED=1 these libraries are
# linked into the C-Chain plugin via github.com/luxfi/chains/evm/cevm
# and become the default execution backend (parallel + GPU EVM).
#
# The release assets live in the PRIVATE lux-private/cevm repo, so the fetch is
# authenticated with the same `ghtok` secret as the private go modules and
# lux-accel above. Best-effort, like lux-accel: only a build that asks for the
# cevm backend (EVM_CGO=1 below) actually needs these.
ARG CGO_ENABLED=1
ARG CEVM_VERSION=v0.51.10
ARG CEVM_REPO=lux-private/cevm
RUN --mount=type=secret,id=ghtok,required=false \
if [ "${CGO_ENABLED}" = "1" ]; then \
ARCH=$(echo ${TARGETPLATFORM} | cut -d / -f2) && \
if [ "$ARCH" = "amd64" ]; then CEVM_ARCH="linux-x86_64"; else CEVM_ARCH="linux-arm64"; fi && \
AUTH=""; [ -s /run/secrets/ghtok ] && AUTH="--header=Authorization: Bearer $(cat /run/secrets/ghtok)"; \
( wget -q ${AUTH:+"$AUTH"} \
"https://github.com/${CEVM_REPO}/releases/download/${CEVM_VERSION}/luxcpp-cevm-${CEVM_ARCH}.tar.gz" \
-O /tmp/cevm.tar.gz \
&& tar -xzf /tmp/cevm.tar.gz -C /usr/local \
&& rm /tmp/cevm.tar.gz \
&& ldconfig 2>/dev/null \
) || echo "WARN: cevm ${CEVM_VERSION} fetch skipped (unreachable; needed only when EVM_CGO=1)" ; \
else \
echo "CGO_ENABLED=0: skipping cevm fetch (pure-Go fallback build)" ; \
fi
# Build node. CGO_ENABLED=1 (default) links luxcpp/cevm for parallel + GPU EVM.
# Set CGO_ENABLED=0 for portable pure-Go builds without the native libs.
ARG RACE_FLAG=""
ARG BUILD_SCRIPT=build.sh
ARG LUXD_COMMIT=""
ENV CGO_ENABLED=${CGO_ENABLED}
RUN . ./build_env.sh && \
# `COPY . .` above restored the committed go.sum (stale first-party hashes when
# a luxfi/hanzoai module was re-tagged). Re-strip first-party lines so -mod=mod
# re-records the CURRENT content hashes already in the module cache (from the
# `go mod download` step). Without this, a re-tag => go.sum SECURITY ERROR.
sed -i -E '/^github.com\/(luxfi|hanzoai)\//d' go.sum && \
echo "{CC=$CC, TARGETPLATFORM=$TARGETPLATFORM, BUILDPLATFORM=$BUILDPLATFORM, CGO_ENABLED=${CGO_ENABLED}}" && \
export GOARCH=$(echo ${TARGETPLATFORM} | cut -d / -f2) && \
export LUXD_COMMIT="${LUXD_COMMIT}" && \
GOFLAGS="-mod=mod" ./scripts/${BUILD_SCRIPT} ${RACE_FLAG}
# ============= EVM Plugin Stage ================
# Build EVM plugin from source (includes custom precompile registry).
# EVM_VERSION must pin a luxfi/evm release whose go.mod points at a
# luxfi/node version that has the runtime.EngineAddressKey rename
# (LUX_VM_RUNTIME_ENGINE_ADDR → VM_RUNTIME_ENGINE_ADDR landed in
# 4ae211d46d on 2026-05-15). v0.18.14 pins node v1.27.6 which is
# post-rename. Older evm tags (e.g. v0.8.40 → node v1.23.4) ship a
# plugin that os.Getenv()'s the old key, mismatching the host that
# sets the new key, and C-chain never bootstraps.
#
# v0.19.0 (Quasar Edition): EtnaTimestamp Go field + json tag renamed
# to QuasarTimestamp/quasarTimestamp. Strict upgradeBytes decode via
# json.NewDecoder(...).DisallowUnknownFields() — stale etnaTimestamp
# in any deployed upgrade.json now fails parse loudly at boot rather
# than silently disabling the fork.
#
# v0.19.1 bumps luxfi/precompile v0.5.27 → v0.5.35 (commit 794912f):
# each of the 7 EIP-2537 bls12381 sub-configs (G1/G2 × {Add,Mul,MSM} +
# Pairing) now returns its own ConfigKey, fixing a Key() collision
# that forced #114 to drop bls12381 entries from mainnet upgrade.json.
# Re-adding them is gated on this plugin version.
#
# v0.19.2 bumps luxfi/vm v1.1.5 → v1.1.6, which drops the obsolete
# GetNetworkUpgrades() method on the VMContext interface. Required by
# the Etna→Quasar rename in v0.19.0: vm v1.1.5 still expected
# upgrade.Config to implement the old IsEtnaActivated() runtime
# interface, which no longer exists.
#
# v0.19.3 closes the cascade: bumps vm v1.1.6 → v1.1.7 + runtime
# v1.0.1 → v1.1.0. runtime v1.1.0 ripped the always-true
# NetworkUpgrades interface (decomplect 9e6e597) and renamed
# XAssetID → UTXOAssetID (refactor 034ec47). v1.1.6 anticipated the
# rip in rpc/context.go but still pinned the old runtime, leaving
# itself internally inconsistent. v1.1.7 pins the post-rip runtime.
# MUST track node's go.mod luxfi/evm (the C-Chain EVM plugin = the native 0x9999
# receipt/atomic surface). v1.99.31 = the native-atomic seam (precompile v0.5.51,
# geth v1.17.12 CallIndex). Bump this with every evm release or the bundled
# C-Chain plugin silently goes stale vs node's deps.
#
# v1.99.33 (precompile v0.5.53): 0x9999 DEX settlement activates at a SINGLE canonical
# dated fork — extras.DexSettleActivationTime = 1766704800 (Dec 25 2025 00:00:00 UTC) —
# with ZERO per-net config (no dexSettleConfig genesis/upgrade entry; one built-in fork,
# identical on every net). At the fork it BOTH enables dispatch AND installs the standard
# EXTCODESIZE marker (nonce=1 + code) into 0x9999 going forward, so eth_getCode(0x9999)
# !=0x and a typed Solidity IPoolManager(0x9999).swap(...) passes the contract-existence
# guard. The marker is installed FORWARD (never in historical genesis), so pre-Dec-25
# history (the ~/work/lux/state RLP snapshot, replayed via admin.importChain) stays
# byte-identical to canonical state — a pre-fork value transfer to 0x9999 hits a PLAIN
# account, not the precompile. The D-Chain (dexvm) peer is resolved at RUNTIME via the
# consensus-context "D" alias (contract.AtomicState.DChainID()) and the
# protocolFeeController is the built-in DAO treasury. 0x9010 is REMOVED (not a registered
# precompile, no dispatch, no forward); 0x9999 is the SOLE canonical DEX precompile. For a
# fresh net whose genesis ts >= the fork, the marker is present from block 0.
#
# v1.99.34 (precompile v0.5.54): fixes a consensus-divergence on the relaunch/replay path.
# The EVM dispatch gate (core.LuxPrecompileOverrider.PrecompileOverride) used to read the
# process-global params.lastRulesContext (via GetRulesExtra(Rules{})), which is rewritten
# last-writer-wins by every concurrent Rules() call (eth_call/estimateGas/worker). On a
# live, RPC-serving post-fork node replaying a PRE-fork RLP block (admin.importChain), a
# concurrent post-fork eth_call could overwrite the global timestamp between the verify
# goroutine's NewEVM(pre-fork block) and its tx dispatch — making the pre-fork block see
# 0x9999 ENABLED and dispatch SettleContract.Run during plain-account execution => state
# divergence / consensus split. Fix: the dispatch gate now decides from the overrider's OWN
# per-EVM fields (o.chainConfig + o.timestamp) via the pure params.GetExtrasRules, never the
# global — so every replay of a block yields the same enabled set on every validator. Also:
# the registry stateDBBridge.SubBalance fallback now FAILS CLOSED (panic → reverted call)
# instead of returning a zero "previous balance" without debiting (silent native mint); and
# the genesis-config builders (SetAllGenesisPrecompiles/AllGenesisPrecompiles) skip AlwaysOn
# modules so 0x9999 can never get a timestamp-0 genesis config that bypasses the dated fork.
# The money path (V4 swap ABI, marker install, two-phase atomic settle) is byte-for-byte
# unchanged: ONLY the dispatch-path timestamp source, the SubBalance fail-mode, the genesis
# builder guard, and stale 0x9010 comments changed.
#
# v1.99.37 (precompile v0.5.57): wires the 0x9999 ERC-20 Call surface to the DEX
# settlement precompile (commit 2cf30e43d) and gates that Call surface to the DEX
# settlement family 0x9999/0x9996 (commit 9579f2e34). Before this, a CALL into
# 0x9999's ERC-20 settle path saw a nil PrecompileEnv (GetPrecompileEnv == nil) and
# could not resolve the token-transfer Call seam — the two-phase atomic settle's
# ERC-20 leg had no env to execute against. precompile v0.5.57 also adds the
# CALL-only DELEGATECALL guard (commit feeaab5a0) so the settle surface is reachable
# only via CALL (not DELEGATECALL, which would run it in the caller's context). Also
# converges deps to latest patch within v1.x.x (threshold v1.9.9, crypto v1.19.21,
# database v1.20.3, geth v1.17.12, warp v1.19.5, vm v1.2.5, api v1.0.15 — the
# UTXOAssetID rename that fixed the LuxAssetID build break) and removes the dead
# vendored dexConfig upgrade fixtures. The 0x9999 swap ABI + dated-fork activation
# (DexSettleActivationTime = 1766704800) are unchanged from v1.99.34.
#
# v1.99.40 (precompile v0.5.59, chains v1.3.19, consensus v1.25.21): the permissionless
# 0x9999 DEX value path lands end-to-end. precompile v0.5.58/59 = AssetResolver (real
# on-chain canonical resolution, NO admin allowlist), one synchronous router/book/journal,
# minOut on every route, no keeper/venue/live-ZAP/second-book; the env→Call ERC-20 vault
# seam + in-state-vault resolution make a real ERC-20 settle. evm v1.99.39 = the
# reprocess-bind fix: NewBlockChain binds the chain Runtime (networkID/C-Chain id) BEFORE
# startup reprocess, so an unclean restart after a 0x9999 swap re-executes the committed
# swap with the correct identity instead of (0, Empty) — previously that reverted the swap,
# failed ValidateState, and BRICKED the node. Proven on-node: real swap → kill -9 →
# clean reboot, state intact. v1.99.40 = v1.99.39 + deps to latest. consensus v1.25.21 =
# stake-weighted alpha-of-K quorum finality + per-height single-finalize + epoch-bound certs.
# v1.104.9 bumps luxfi/api v1.0.15 -> v1.0.16 (indirect via luxfi/vm). v1.0.16
# APPENDED InitializeResponse.Capabilities (uint64) for the Quasar-export
# handshake (api commit 1f2dc5a). The node pins api v1.0.16 and DECODES that
# field; an EVM plugin built at v1.104.8 (api v1.0.15) omits it, so the node's
# strict struct decode hits "zap decode initialize response: unexpected EOF"
# and EVERY EVM chain (C + L2s hanzo/zoo/pars/spc, all mgj786) fails to
# initialize. The api bump is code-free for plugins (chains v1.7.4->v1.7.5
# adopted it with a go.mod-only diff), so v1.104.9 = v1.104.8 + api alignment.
# C-Chain execution backend. The default is the pure-Go EVM, which is what
# every image has shipped so far. EVM_CGO=1 EVM_TAGS=cevm links luxcpp/cevm
# instead — that pair is what makes AutoEVM resolve to CppEVM. The libraries
# come from the cevm fetch in this same stage above.
ARG EVM_CGO=0
ARG EVM_TAGS=""
# v1.104.22 realigns the plugin with THIS node's own go.mod: api v1.0.16 ->
# v1.1.1, vm v1.2.6 -> v1.3.1, geth v1.17.12 -> v1.20.1. Node main pins api
# v1.1.1, so pinning an evm at api v1.0.16 reintroduces the exact
# InitializeResponse decode mismatch described above, just in the opposite
# direction — keep this ARG and node's go.mod on the same api/vm/geth line.
#
# v1.104.14 is also the FIRST evm tag carrying the C-Chain fee-split seam
# (core/fee_split.go creditTxFee + extras.FeeSplitTimestamp/FeeRewardVault).
# Below it, encoding/json silently DISCARDS the genesis "feeSplitTimestamp"
# key: a chain configured for the 50/50 split instead routes 100% of every fee
# to the block coinbase and the reward vault 0x0100..0002 stays 0 forever, with
# nothing burned. The split stays dormant wherever feeSplitTimestamp is absent
# (mainnet), so this bump is behaviour-preserving there.
ARG EVM_VERSION=v1.104.49
ARG EVM_VM_ID=mgj786NP7uDwBCcq6YwThhaN8FLyybkCa4zBWTQbNgmK6k9A6
# the pinned evm go.mod may pin a dead luxfi/upgrade pseudo-version
# (v1.0.1-0.20260603055252-f51810805436 — commit pruned from origin). Heal it to
# the released upgrade v1.0.1 tag (semver-forward, not a downgrade). Then strip
# first-party go.sum lines so -mod=mod re-records current content hashes for the
# re-published luxfi modules at their pinned versions (integrity repair, no version drift).
RUN --mount=type=cache,target=/root/.cache/go-build \
mkdir -p /luxd/build/plugins && \
git clone --depth 1 --branch ${EVM_VERSION} https://github.com/luxfi/evm.git /tmp/evm && \
cd /tmp/evm && \
. /build/build_env.sh && \
go mod edit -require=github.com/luxfi/upgrade@v1.0.1 && \
# evm v1.99.52 = v1.99.51 + the idle-builder bounded-wake backstop
# (startPendingTxPoll: a 500ms mempool re-poll so a tx after idle wakes the
# builder within a bounded window — closes the lost-wakeup/subscribe-gap;
# deterministic, wake-timing only). v1.99.51 = the block-production stall fix
# (WaitForEvent builder-ready race + target-rate build pacing — un-freezes the
# C-Chain that stalled at the imported frontier) on top of precompile v0.16.0
# enable-everything
# (wallet curves + standard precompiles enabled; fflonk fail-closed; accel
# byte-identity; DEX big.Rat + determinism). Pin chains v1.4.8 (warp
# consolidated to one luxfi/warp helper; graphvm genesis-last-accepted fix)
# to match the chain-VM plugin stage (CHAINS_REF) below.
go mod edit -require=github.com/luxfi/chains@v1.7.0 && \
find /tmp/evm -name go.sum -exec sed -i -E '/^github.com\/(luxfi|hanzoai)\//d' {} + && \
GOARCH=$(echo ${TARGETPLATFORM} | cut -d / -f2) \
CGO_ENABLED=${EVM_CGO} GOFLAGS=-mod=mod \
go build -ldflags="-s -w" -tags "${EVM_TAGS}" -o /luxd/build/plugins/${EVM_VM_ID} ./plugin && \
# Ship the fail-closed offline C-Chain repair tool from the exact same EVM
# source as the plugin. It is inert during normal node startup; operators can
# run it from an init container while luxd is stopped to reconcile a legacy
# pre-certification accepted tip without introducing a second image or an
# unpinned recovery binary.
GOARCH=$(echo ${TARGETPLATFORM} | cut -d / -f2) \
CGO_ENABLED=0 GOFLAGS=-mod=mod \
go build -ldflags="-s -w" -o /luxd/build/repair-cchain ./cmd/repair-cchain && \
chmod +x /luxd/build/plugins/${EVM_VM_ID} && \
chmod +x /luxd/build/repair-cchain && \
# SEAM PIN ASSERTION — the C-Chain half. See the D-Chain twin at the dex build.
#
# The node serves each chain's shared memory to its plugin, so SharedMemory here is
# the ZAP client, and atomiczap.Apply REFUSES a database batch outright
# (ErrBatchUnsupported). A plugin pinned below the batch-less Accept therefore turns
# every block that stages an atomic op into a FATAL Accept — and Accept failure is
# fail-closed in the engine ("refusing to advance finality past the VM's applied
# state"), so the first 0x9999 settlement would STOP THE CHAIN. That is exactly what
# v1.36.60 shipped: shared memory served, plugins pinned pre-fix.
#
# This is a BUILD CHECK, not a runtime gate. It cannot skip, defer or disable
# anything; it only refuses to produce an image whose pins contradict the transport.
# Positive signal AND negative signal, so neither a rename nor a deletion passes
# silently.
grep -qE 'atomicSM\.Apply\(atomicReqs\)' plugin/evm/block.go \
|| { echo "FATAL: evm ${EVM_VERSION} plugin/evm/block.go has no batch-less 'atomicSM.Apply(atomicReqs)'. A pin below the fix halts the chain on the first atomic op."; exit 1; } && \
! grep -nE '\.Apply\([^)]+,' plugin/evm/block.go \
|| { echo "FATAL: evm ${EVM_VERSION} passes a second argument to Apply — atomiczap refuses a batch across the process boundary (ErrBatchUnsupported)."; exit 1; } && \
rm -rf /tmp/evm
# ============= Chain VM Plugin Stage ================
# Build the 10 chain VM plugins that github.com/luxfi/chains implements.
#
# The D-Chain is NOT among them. Its slot is filled by luxfi/dex cmd/dexd in the
# next stage, so dexvm is absent from both the table and the build list below. A
# slot holds one binary: adding a D build here would not add a fallback, it would
# race the matcher and win or lose by ordering.
#
# VM ID table (CB58-encoded ids.ID byte arrays from each factory.go):
# aivm -> juFxSrbCM4wszxddKepj1GWwmrn9YgN1g4n3VUWPpRo9JjERA
# bridgevm -> kMhHABHM8j4bH94MCc4rsTNdo5E9En37MMyiujk4WdNxgXFsY
# graphvm -> nZQm4Dmg1rjX18rb8maL9gamYyXPf1xCvF7ymWzxp6a1nSQTt
# identityvm -> oR6tnZHezwogyf9fRnomNXC9ojwCEBAU6jdUzpgy2PB1tD7fM
# keyvm -> pJJCSV7hHYVY6TUZwR8qUPAfuhX8JLb2C1AzNSezrYNbgau8M
# oraclevm -> r5m1ujrmXxVcQetG3CQfuDLHp2RHKh6vCDaFgBRQfUcTZh7eS
# quantumvm -> ry9Sg8rZdT26iEKvJDmC2wkESs4SDKgZEhk5BgLSwg1EpcNug
# relayvm -> sP6dLqrrBR9w3soP18fbJ3YzZecZdD7DDdfH2cFhhLq7Hy9bz
# mpcvm -> qCURact1n41FcoNBch8iMVBwc9AWie48D118ZNJ5tBdWrvryS
# zkvm -> vv3qPfyTVXZ5ArRZA9Jh4hbYDTBe43f7sgQg4CHfNg1rnnvX9
# This ref MUST equal node's go.mod luxfi/chains pin. go.mod is the authority;
# naming the release's contents here is what let the two drift apart unnoticed.
#
# The plugins are baked into the image and nothing else delivers them, so a ref
# behind go.mod ships VMs the host node was not built against and no operator
# action can correct it short of a new image. It reached eleven tags behind: the
# mpcvm plugin was built from v1.7.6, which predates M-Chain's state and custody
# implementations entirely and still declared the retired `thresholdvm` VMID.
# That last part stayed harmless only because the node resolves a plugin by its
# FILENAME (vms/registry/registry.go), never by the ID the binary declares.
# READ from go.mod rather than restate it. The comment above already said go.mod
# was the authority and that naming the ref here is what let them drift — and it
# drifted anyway, to eleven tags at the worst and to five security commits at the
# time this was replaced. A rule written in a comment beside the value it governs
# is not a rule; deriving the value is.
RUN --mount=type=cache,target=/root/.cache/go-build \
CHAINS_REF="$(awk '$1=="github.com/luxfi/chains"{print $2; exit}' /build/go.mod)" && \
if [ -z "${CHAINS_REF}" ]; then echo "go.mod names no luxfi/chains version" >&2; exit 1; fi && \
echo "building chain VM plugins from go.mod pin ${CHAINS_REF}" && \
git clone --depth 1 --branch ${CHAINS_REF} https://github.com/luxfi/chains.git /tmp/chains && \
find /tmp/chains -name go.sum -exec sed -i -E '/^github.com\/(luxfi|hanzoai)\//d' {} +
# F-Chain ships as the SAME mpcvm binary under a second filename. mpcvm backs
# both surfaces -- M-Chain's MPC signing and F-Chain's FHE compute (mpcvm
# service.go: "the one VM backs all three surfaces"; LP-8200: "F-Chain runs on
# the shared mpcvm substrate"). The node resolves a plugin by FILENAME, never by
# the id the binary declares, so installing it under M's id alone is what made
# F-Chain report "VM plugin not loaded" and opt out on every node of every
# network. It is copied rather than built twice so the two names stay
# byte-identical: one implementation, two ids, not two builds that can drift.
#
# Each VM is an independent Go module under /tmp/chains/<vm>/.
# Build each plugin binary and place it at /luxd/build/plugins/<cb58-vm-id>.
#
# The recursive go.sum strip above lets -mod=mod re-record current content hashes
# for re-published first-party modules at their pinned versions (integrity repair).
# At a tagged CHAINS_REF (v1.3.11) the sibling go.mod replace directives that
# affect `main` are absent, so every VM module builds in isolation. The bridgevm
# plugin (B-Chain) MUST build — it blank-imports crypto/threshold/bls to register
# the BLS threshold scheme; a miss reintroduces the v1.30.16 B-Chain init failure.
RUN --mount=type=cache,target=/root/.cache/go-build \
. /build/build_env.sh && \
export GOARCH=$(echo ${TARGETPLATFORM} | cut -d / -f2) && \
mkdir -p /luxd/build/plugins && \
( cd /tmp/chains/aivm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/juFxSrbCM4wszxddKepj1GWwmrn9YgN1g4n3VUWPpRo9JjERA ./cmd/plugin ) || echo "WARN: aivm plugin build skipped" ; \
( cd /tmp/chains/bridgevm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/kMhHABHM8j4bH94MCc4rsTNdo5E9En37MMyiujk4WdNxgXFsY ./cmd/plugin ) ; \
( cd /tmp/chains/graphvm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/nZQm4Dmg1rjX18rb8maL9gamYyXPf1xCvF7ymWzxp6a1nSQTt ./cmd/plugin ) || echo "WARN: graphvm plugin build skipped" ; \
( cd /tmp/chains/identityvm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/oR6tnZHezwogyf9fRnomNXC9ojwCEBAU6jdUzpgy2PB1tD7fM ./cmd/plugin ) || echo "WARN: identityvm plugin build skipped" ; \
( cd /tmp/chains/keyvm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/pJJCSV7hHYVY6TUZwR8qUPAfuhX8JLb2C1AzNSezrYNbgau8M ./cmd/plugin ) || echo "WARN: keyvm plugin build skipped" ; \
( cd /tmp/chains/oraclevm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/r5m1ujrmXxVcQetG3CQfuDLHp2RHKh6vCDaFgBRQfUcTZh7eS ./cmd/plugin ) || echo "WARN: oraclevm plugin build skipped" ; \
( cd /tmp/chains/quantumvm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/ry9Sg8rZdT26iEKvJDmC2wkESs4SDKgZEhk5BgLSwg1EpcNug ./cmd/plugin ) || echo "WARN: quantumvm plugin build skipped" ; \
( cd /tmp/chains/relayvm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/sP6dLqrrBR9w3soP18fbJ3YzZecZdD7DDdfH2cFhhLq7Hy9bz ./cmd/plugin ) || echo "WARN: relayvm plugin build skipped" ; \
( cd /tmp/chains/mpcvm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/qCURact1n41FcoNBch8iMVBwc9AWie48D118ZNJ5tBdWrvryS ./cmd/plugin ) || echo "WARN: mpcvm plugin build skipped" ; \
( cp /luxd/build/plugins/qCURact1n41FcoNBch8iMVBwc9AWie48D118ZNJ5tBdWrvryS \
/luxd/build/plugins/n6sSsSfbpQBrU9sY4R29U6z8VrmnTo2CntW6da4rRS7qmnGdv ) || echo "WARN: fhevm install skipped" ; \
( cd /tmp/chains/zkvm && CGO_ENABLED=0 GOFLAGS=-mod=mod go build -ldflags="-s -w" \
-o /luxd/build/plugins/vv3qPfyTVXZ5ArRZA9Jh4hbYDTBe43f7sgQg4CHfNg1rnnvX9 ./cmd/plugin ) || echo "WARN: zkvm plugin build skipped" ; \
( chmod +x /luxd/build/plugins/* 2>/dev/null || true ) && \
for p in \
juFxSrbCM4wszxddKepj1GWwmrn9YgN1g4n3VUWPpRo9JjERA \
kMhHABHM8j4bH94MCc4rsTNdo5E9En37MMyiujk4WdNxgXFsY \
nZQm4Dmg1rjX18rb8maL9gamYyXPf1xCvF7ymWzxp6a1nSQTt \
oR6tnZHezwogyf9fRnomNXC9ojwCEBAU6jdUzpgy2PB1tD7fM \
pJJCSV7hHYVY6TUZwR8qUPAfuhX8JLb2C1AzNSezrYNbgau8M \
r5m1ujrmXxVcQetG3CQfuDLHp2RHKh6vCDaFgBRQfUcTZh7eS \
ry9Sg8rZdT26iEKvJDmC2wkESs4SDKgZEhk5BgLSwg1EpcNug \
sP6dLqrrBR9w3soP18fbJ3YzZecZdD7DDdfH2cFhhLq7Hy9bz \
qCURact1n41FcoNBch8iMVBwc9AWie48D118ZNJ5tBdWrvryS \
n6sSsSfbpQBrU9sY4R29U6z8VrmnTo2CntW6da4rRS7qmnGdv \
vv3qPfyTVXZ5ArRZA9Jh4hbYDTBe43f7sgQg4CHfNg1rnnvX9 ; do \
test -s /luxd/build/plugins/$p \
|| { echo "FATAL: required chain-VM plugin $p missing/empty — its build failed above (see the matching WARN line); the runtime-stage hard COPY would otherwise fail cryptically. Surface & fix the real Go build error, or remove the plugin from BOTH the build list and the runtime COPY."; exit 1; } ; \
done && \
rm -rf /tmp/chains
# ============= Native D-Chain DEX VM Plugin Stage ================
# The D-Chain (dexvm slot, vmID mDVT5EWMumBp3LCqvKwuyZQeY1VXr1jvjGNAt8nL4UFiXvqXr)
# is the NATIVE consensus matcher VM: github.com/luxfi/dex/pkg/dchain implements
# block.ChainVM and runs the lx.OrderBook matcher INSIDE luxd consensus
# (Block.Verify against a versiondb overlay; Block.Accept commits) — the trade IS
# the D-Chain state transition, sequenced by luxd's multi-validator engine. This
# REPLACES the chains/dexvm proxy (which relayed clob_* over ZAP to a standalone
# dchain-venue): there is no DexZapEndpoint and no standalone venue in the trading
# path. That proxy is deleted — it kept building for this slot after losing it, and
# its clob_* wire had stopped matching the venue's dex_* names, so installing it
# produced a node that loaded, passed health, and answered nothing.
#
# cmd/dexd (renamed from cmd/dchain in the v1.14 line — it is the ONE D-Chain
# binary, plugin mode by default) wraps the VM in the SAME rpc.Serve plugin
# harness luxfi/evm boots through, and is pure-Go (CGO=0) — the optional GPU AMM
# accelerator in pkg/lx is a separate concern gated by its own cuda/metal tags and
# is NOT linked here. v1.5.10 is the first tag whose cmd/dchain builds CGO=0
# (drops the phantom dchain+cgo gate); v1.5.11 wires CLOB order ingestion over the
# node HTTP router (VM.CreateHandlers -> /v1/bc/D/dex/<method>, pkg/dchain/ingest.go)
# so an order POSTed to the node flows submitTx -> mempool -> consensus -> Verify
# (match) -> Accept; v1.5.12 persists the head block so the VM survives a restart
# once advanced past genesis (GetBlock(lastAccepted) no longer ErrNotFound);
# v1.5.13 indexes processing blocks so the plugin transport's ID-only Accept
# (GetBlock(builtID)) resolves the just-built block — without it the engine's
# self-finalize Accept is a silent no-op and the clob submitTx waiter hangs (no
# D-Chain blocks). v1.5.14 consumes luxfi/database v1.20.4, which fixes prefixdb
# returning ZERO rows for a nil-start prefix scan over a prefixdb-wrapped chain
# DB (every plugin VM via vm/rpc) — that stranded the native D-Chain order book
# (rebuildBookFromDB folded empty -> 0 fills despite committed asks). v1.5.15 adds
# the committed-state READ surface (clob_get_trades/orders/markets/book over
# /v1/bc/D/dex/<method>, pkg/dchain/read.go): read-only JSON of the durable trade
# log / resting book / markets, served beside the writes with ZERO consensus
# impact. Needed to VERIFY a fill replicated identically across validators (query
# every node, diff the trade rows + head root) and to feed markets-display (native
# fills are trade: rows). Bump with every dex release that changes the VM, like
# CHAINS_REF for the other 10 VMs.
#
# v1.5.15 -> v1.14.4 IS A LINEAGE CHANGE, NOT A PATCH. v1.5.15 is not an ancestor
# of main (278 commits diverged) and its pkg/dchain has NO atomic.go and NO
# drive.go: that D-Chain cannot import a C->D intent or export a D->C fill at
# all. Every fleet running it has a structurally absent settlement seam, so
# fixing the ZAP atomic transport and the intent discovery trait changes nothing
# while this pin stands. The v1.14 line is where the seam lives, and v1.14.4
# pins luxfi/vm v1.3.6 so the plugin actually receives shared memory.
#
# Treat this as a D-Chain VM upgrade and roll it like one: it changes block
# production (the autonomous seam drive emits import/export txs inside
# BuildBlock), so a mixed fleet across this line does NOT agree on block
# contents. One validator at a time, D-Chain only, and never below quorum.
#
# v1.14.4 -> v1.14.6 IS A WIRE FLAG DAY, and two separate things ride on it.
#
# (1) THE SEAM ONLY NOW PLACES AN ORDER. Through v1.14.5 the C->D object carried
# VALUE only, so the D-Chain consumed an intent, credited the taker, and stopped.
# Nothing crossed, nothing was exported, and 0x9999 emitted zero events on every
# chain since genesis. v1.14.6 carries the taker's operation (market, side,
# limit, size) in the object and builds their order from it. It REQUIRES
# precompile >= v0.19.7 on the C side (hence EVM_VERSION v1.104.26): the two
# repos pin the same 118-byte golden vector and the same .v2 discovery trait.
#
# (2) THE TxImport BODY WIDENED, 32 -> 150 bytes, because the object's own bytes
# now ride in the transaction (execution that reads shared memory cannot be
# replayed: the consuming Remove lands at Accept, so the next node to sync finds
# the object gone and dies on `execution root mismatch`). An OLD node's
# bodySize(TxImport) is 32, so a new seam block fails ParseBlock outright — it is
# rejected as UNPARSEABLE, never executed differently. A mixed fleet across this
# line STALLS on seam blocks; it cannot fork. No historical block on any network
# contains a TxImport, so there is no retroactive effect and no re-sync risk
# below the line.
#
# Roll every D validator to this image BEFORE the first v2 intent is minted on C.
# Until then the drive finds nothing (it enumerates the .v2 trait, which no
# pre-v0.19.7 C node writes) and emits no seam tx at all.
#
# THE DRIVE IS STATUS-DRIVEN as of v1.14.10, and that is a correctness fix, not a
# refactor: through v1.14.9 the drive emitted the import and the taker's order as a PAIR
# in one block, so a proposer that included the import and omitted the order produced a
# valid block whose escrow was stuck open forever — the taker's value stranded in a
# keyless D account while C's deadline reclaim refunded them. Any proposer could do it to
# any swap, for free. Do not pin a DEX_REF below v1.14.10 with the seam enabled.
#
# THE DEPLOYED FLEET IS NOT ON THE PIN ABOVE IT. Probing devnet's running plugin
# (grep the vendored module strings out of /data/plugins/mDVT5...) shows luxfi/dex
# v1.5.15, not v1.14.x — the image predates this ARG. That matters twice:
#
# - v1.5.15 HAS the HTTP surface, under the OLD method names: /v1/bc/D/dex/clob_*
# answers 200 today (clob_get_markets returns height 0, markets []). A probe of
# /v1/bc/D/dex/dex_* returning 404 means "old names", NOT "no surface". The
# rename landed in dex 28970d8 (clob_* -> dex_*).
# - v1.5.15 has NO atomic.go and NO drive.go, so that D-Chain cannot import a C->D
# intent or export a D->C fill AT ALL. Every fleet on it has a structurally
# absent seam, which is the real reason nothing has ever traded — not the wire,
# not the trait, not funding.
#
# So this bump is a D-Chain LINEAGE change on the running fleet (v1.5.15 -> v1.14.x),
# not a patch. Devnet's D-Chain is at height 0 with 0 markets on all 5 validators, so
# there is no accumulated state for it to disagree with.
# v1.14.31, not v1.14.25. Two things landed in between that a fleet must not run
# without:
#
# 1. executeImport guarded only the C chain id while executeExport also guarded
# shared memory. The halves disagreed, and Block.verifyImports returned nil
# when shared memory was nil on the reasoning that "executeImport rejects
# without a C chain id" — it rejects without a chain ID, not without shared
# memory. Both ends guard both conditions now.
#
# 2. A duplicate delivery of an already-consumed claim read as unbacked, so the
# block was rejected, Reject requeued every tx, and the next build failed
# identically — a permanent halt, measured at four rounds stuck on one
# height. Delivery is permissionless by design, so duplicates are the steady
# state rather than an attack. verifyImports rejects the block for safety;
# BuildBlock holds the delivery and retries for liveness.
#
# The genesis construction is byte-identical across that range — pkg/dchain
# genesis.go and state.go do not differ between v1.14.25 and v1.14.31, and all
# twelve genesis vectors including TestGenesisLiveFleets pass at v1.14.31. So the
# founding values published for the re-foundation are unchanged by this bump.
ARG DEX_REF=v1.14.31
RUN --mount=type=cache,target=/root/.cache/go-build \
git clone --depth 1 --branch ${DEX_REF} https://github.com/luxfi/dex.git /tmp/dex && \
find /tmp/dex -name go.sum -exec sed -i -E '/^github.com\/(luxfi|hanzoai)\//d' {} + && \
cd /tmp/dex && \
# dex v1.14.4 pins luxfi/api v1.1.3 and luxfi/vm v1.3.6 itself, so the ZAP
# wire aligns with the node without an override here. The former forced
# api@v1.0.16 is GONE: forcing an api version below what luxfi/vm requires
# would strip the cross-chain atomic messages back out of the dexvm plugin
# and re-dark the seam this bump exists to open.
. /build/build_env.sh && \
GOARCH=$(echo ${TARGETPLATFORM} | cut -d / -f2) \
CGO_ENABLED=0 GOFLAGS=-mod=mod \
go build -ldflags="-s -w" \
-o /luxd/build/plugins/mDVT5EWMumBp3LCqvKwuyZQeY1VXr1jvjGNAt8nL4UFiXvqXr ./cmd/dexd && \
chmod +x /luxd/build/plugins/mDVT5EWMumBp3LCqvKwuyZQeY1VXr1jvjGNAt8nL4UFiXvqXr && \
test -s /luxd/build/plugins/mDVT5EWMumBp3LCqvKwuyZQeY1VXr1jvjGNAt8nL4UFiXvqXr \
|| { echo "FATAL: native D-Chain (dexvm) plugin missing — D-Chain cannot start"; exit 1; } && \
# SEAM PIN ASSERTION — the D-Chain half. Twin of the C-Chain check at the evm build;
# the same ErrBatchUnsupported refusal halts D on its first seam settle.
grep -qE 'sm\.Apply\(ar\.reqs\)' pkg/dchain/atomic.go \
|| { echo "FATAL: dex ${DEX_REF} pkg/dchain/atomic.go has no batch-less 'sm.Apply(ar.reqs)'. A pin below v1.14.5 halts D-Chain on the first seam settle."; exit 1; } && \
! grep -nE '\.Apply\([^)]+,' pkg/dchain/atomic.go \
|| { echo "FATAL: dex ${DEX_REF} passes a second argument to Apply — atomiczap refuses a batch across the process boundary (ErrBatchUnsupported)."; exit 1; } && \
# The D-side reproducibility defect: executeImport must take the object as a
# PARAMETER (it travels in the tx body), never read it from shared memory during
# execute(). execute() is compared against the block's execRoot, and sm.Get reads
# live cross-chain memory outside that root — a node re-executing an accepted block
# finds the object already consumed, derives a different root, and can never sync
# past the first swap the seam settles. Fixed in v1.14.10; pinned here so it stays fixed.
grep -qE 'func \(vm \*VM\) executeImport\(.*object \[\]byte' pkg/dchain/atomic.go \
|| { echo "FATAL: dex ${DEX_REF} executeImport does not carry the object as a parameter — a re-executing node will fail 'execution root mismatch' and can never bootstrap past a settle."; exit 1; } && \
# WIRE NAMESPACE ASSERTION. A VM is reachable only through the method names it
# answers, and a plugin that answers names nobody dials still loads and still
# passes health — every order is a silent no-op on a chain that reports up. The
# venue registers dex_*; clob_* was the wire of a proxy (formerly chains/dexvm)
# that spoke to nothing and has been deleted. Checked here because this is where
# the source exists; no runtime probe distinguishes the two.
grep -rq '"dex_place"' pkg/ \
|| { echo "FATAL: dex ${DEX_REF} does not register the dex_* wire namespace — a D-Chain plugin built from it would load into the slot and answer nothing."; exit 1; } && \
! grep -rq '"clob_' pkg/ \
|| { echo "FATAL: dex ${DEX_REF} carries the dead clob_* wire namespace. Nothing answers those names on either side."; exit 1; } && \
# GENESIS ASSERTION. The D-Chain's height-0 block must come from the document
# recorded in its P-Chain CreateChainTx, never from this binary. A pin that
# derives genesis from compiled-in defaults gives every wiped node a chain of
# its own, silently and while reporting healthy — and wiping chainData is the
# ordinary repair, so the ordinary repair is what strands the validator. Both
# halves are required: read the delivered document, and refuse to found from
# one that never arrived. Asserted here because this line selects the source
# and no runtime probe distinguishes a right genesis from a lucky one.
grep -q 'refusing to found a chain with no creation document' pkg/dchain/genesis.go \
|| { echo "FATAL: dex ${DEX_REF} founds a chain from an empty creation document — a wiped node would adopt this binary's genesis and partition alone. Pin v1.14.25 or later."; exit 1; } && \
rm -rf /tmp/dex
# lpm (Lux Plugin Manager) -- optional, skip if build fails
RUN --mount=type=cache,target=/root/.cache/go-build \
--mount=type=cache,target=/root/go/pkg/mod \
GOARCH=$(echo ${TARGETPLATFORM} | cut -d / -f2) && \
git clone --depth 1 https://github.com/luxfi/lpm.git /tmp/lpm && \
cd /tmp/lpm && \
CGO_ENABLED=0 go build -ldflags="-s -w" -o /luxd/build/lpm ./main && \
rm -rf /tmp/lpm || echo "WARN: lpm build skipped (non-critical)"
# Create this directory in the builder to avoid requiring anything to be executed in the
# potentially emulated execution container.
RUN mkdir -p /luxd/build
# ============= Runtime Stage ================
# Commands executed in this stage may be emulated (i.e. very slow) if TARGETPLATFORM and
# BUILDPLATFORM have different arches.
FROM debian:12-slim AS execution
# Install runtime dependencies (curl for RPC, git for lpm source installs, ca-certificates for TLS)
RUN apt-get update && apt-get install -y --no-install-recommends \
curl ca-certificates git \
&& rm -rf /var/lib/apt/lists/*
# GPU crypto library (optional -- only present when built with CGO_ENABLED=1 + luxcpp).
# Pure Go fallbacks are used when the library is absent.
RUN ldconfig 2>/dev/null || true
# Maintain compatibility with previous images.
# In the builder stage /luxd/build contains ONLY plugins/ (the luxd + lpm binaries
# live at /build/build and are COPYed separately below). The plugin set (~192MB of
# 12 VM plugins) is split across several COPY layers so each blob stays well under
# ~100MB: a single monolithic plugins layer cannot reliably complete its registry
# blob upload over a contended uplink, whereas sub-100MB layers push reliably (and
# resume independently by digest).
# plugin group 1
COPY --from=builder \
/luxd/build/plugins/mgj786NP7uDwBCcq6YwThhaN8FLyybkCa4zBWTQbNgmK6k9A6 \
/luxd/build/plugins/juFxSrbCM4wszxddKepj1GWwmrn9YgN1g4n3VUWPpRo9JjERA \
/luxd/build/plugins/kMhHABHM8j4bH94MCc4rsTNdo5E9En37MMyiujk4WdNxgXFsY \
/luxd/build/plugins/
# plugin group 2
COPY --from=builder \
/luxd/build/plugins/mDVT5EWMumBp3LCqvKwuyZQeY1VXr1jvjGNAt8nL4UFiXvqXr \
/luxd/build/plugins/nZQm4Dmg1rjX18rb8maL9gamYyXPf1xCvF7ymWzxp6a1nSQTt \
/luxd/build/plugins/oR6tnZHezwogyf9fRnomNXC9ojwCEBAU6jdUzpgy2PB1tD7fM \
/luxd/build/plugins/pJJCSV7hHYVY6TUZwR8qUPAfuhX8JLb2C1AzNSezrYNbgau8M \
/luxd/build/plugins/
# plugin group 3
COPY --from=builder \
/luxd/build/plugins/r5m1ujrmXxVcQetG3CQfuDLHp2RHKh6vCDaFgBRQfUcTZh7eS \
/luxd/build/plugins/ry9Sg8rZdT26iEKvJDmC2wkESs4SDKgZEhk5BgLSwg1EpcNug \
/luxd/build/plugins/sP6dLqrrBR9w3soP18fbJ3YzZecZdD7DDdfH2cFhhLq7Hy9bz \
/luxd/build/plugins/qCURact1n41FcoNBch8iMVBwc9AWie48D118ZNJ5tBdWrvryS \
/luxd/build/plugins/vv3qPfyTVXZ5ArRZA9Jh4hbYDTBe43f7sgQg4CHfNg1rnnvX9 \
/luxd/build/plugins/
# plugin group 4 — F-Chain. Same bytes as mpcvm in group 3, under F's id,
# because one binary serves both surfaces and the registry resolves by filename.
# It needs its own COPY: the builder stage writes it, but only files named here
# reach the final image, which is why building it was not enough to ship it.
COPY --from=builder \
/luxd/build/plugins/n6sSsSfbpQBrU9sY4R29U6z8VrmnTo2CntW6da4rRS7qmnGdv \
/luxd/build/plugins/
WORKDIR /luxd/build
# Copy the executables into the container
COPY --from=builder /build/build/ .
COPY --from=builder /luxd/build/repair-cchain ./repair-cchain
# Create plugins directory and lpm state directory
RUN mkdir -p /luxd/build/plugins /root/.lpm /root/.lux/plugins
# Add lpm to PATH
ENV PATH="/luxd/build:${PATH}"
EXPOSE 9630 9631
CMD [ "./luxd" ]