[marceloceccon]

Chain cut in merchant-sdk: Solana archived, Polygon dropped.

⚠️ This is a breaking API change, and the version is deliberately NOT bumped

Removed from the public surface: the ./solana subpath, SolanaCommitmentLevel,
ConfirmationPolicy.commitmentLevel(), createHighValueConfirmationPolicy(),
SolanaGasBreakdown, and 'solana' from the ChainFamily / PaymentFamily /
TelemetryChain unions.

I removed those rather than leaving them returning null for every chain. A policy
that still advertises a commitment vocabulary no chain uses is a trap for the next
integrator — and the work pack is explicit that dead toggles are not acceptable.

package.json still says 0.5.0 on purpose: ci.yml publishes
<version>-dev.<run> to GitHub Packages on every push to develop, so bumping
here would ship a breaking prerelease off the merge. The 0.6.0 release decision
is yours — humanpending.md HP-3.

What changed

src/solana/ (11 files) and the two Solana specs moved to archive/solana/. They
are inert: tsconfig includes only src, ESLint globs only src/**, the Vite
entry map is explicit, and I added a test.exclude for archive/** to
vitest.config.ts — without it Vitest's default glob would still have collected
the archived spec. Six @solana/* dependencies gone; lockfile −12.8k lines.

Polygon: chainId 137 out of DEFAULT_CHAINS, CHAIN_ICONS,
COINGECKO_CHAIN_IDS, KNOWN_CONTRACT_ADDRESSES,
DEFAULT_CONFIRMATION_THRESHOLDS, CHAIN_FAMILY_REGISTRY,
DEFAULT_SECONDS_TO_FINALITY, the price-feed matic-network fallback, the wagmi
chain list, and both EVM chain-id allowlists.

New drop guards in confirmationPolicy.test.ts: 137 and the Solana sentinels
900–902 must be absent from every registry, must resolve as unknown chains (12
confirmations, 0s ETA) rather than as their old selves, and no registry entry may
claim a family outside evm｜tron. Re-adding a dropped chain without a founder
decision now fails the suite.

Four fixes that are NOT about chains — please read

npm run typecheck could not run at all before this PR, which meant the cut was
unverifiable. Fixed minimally:

1. tsconfig.json — dropped the deprecated baseUrl and the @/* paths
   alias it enabled (nothing imported it). TypeScript 6 makes baseUrl a hard
   error, so tsc aborted before reading a single source file.
2. src/hooks/useWallet.ts — wagmi v3's useBalance no longer returns
   formatted; now formats value/decimals with viem's formatUnits. Public
   balance: string | null contract unchanged. This was a real compile error in
   mainline.
3. src/css.d.ts (new) — TS6 needs a declaration for the side-effect
   import './styles.css'.
4. .npmrc (new) — legacy-peer-deps=true. eslint-plugin-jsx-a11y@6.10.2
   (its newest release) peers eslint ^3…^9 while this package pins ^10, so a
   clean npm ci fails ERESOLVE. The committed lockfile was already built that
   way; this makes it explicit.

If you'd rather these landed separately, they are the only non-chain changes in the
whole nine-repo set — easy to split.

Verification — executed

• npm ci → clean install, 465 packages (previously failed ERESOLVE).
• npm run typecheck → PASS.
• npm test → 165 passed, 4 failed. The same 4 permit.test.ts cases fail
  identically on the pre-cut tree — I verified by stashing the cut and re-running.
  They hit ADR-0004's KNOWN_PERMIT_TOKENS gate with a non-allow-listed token
  domain, and the shipped allowlist holds one placeholder digest. Making them pass
  means deciding the real allowlist contents, not editing a test
  (humanpending.md HP-5). The cut introduces no new failures.

Still red, pre-existing, not fixed here (humanpending.md HP-4): npm run lint
crashes because eslint-plugin-react is not ESLint-10 compatible, and
npm run build's Vite half fails on Tailwind v4's moved PostCSS plugin (and then
on a missing esbuild). I attempted the Tailwind fix, hit the second failure, and
reverted rather than ship an unverifiable dependency change inside a chain PR.

---

Locked decisions applied

#	Decision	Status
D1	Drop Polygon permanently from the mainline product	LOCKED — applied
D2	Archive Solana out of mainline (quarantine, not delete)	LOCKED — applied
D3	Active chain families: EVM (non-Polygon) + TRON	LOCKED — applied
D4	GTM needs ≥4 mainnets; in-tree after the cut = Ethereum + Base + TRON = 3	BLOCKED on D4a
D4a	Mainnet #4 = _________	OPEN — deliberately not invented
D5	First money stays testnet → canary → the four mainnets together; this pack is product/matrix cleanup, not deploy	LOCKED — respected

No fourth chain is named anywhere in this PR set. Chain matrices carry an explicit
TODO(D4a) and no placeholder row appears in any catalog, seed or UI.

Inventory summary (WI-0)

Nine of ten repos needed changes. portal-shared needed none — its only
apparent hit was matic inside the word automatic, so no branch and no PR exist
for it.

Worth knowing about the scan: the work pack's regex over-matches badly (matic →
automatic, anchor → HTML anchors, spl → split), and the monorepo
.gitignore ignores every sub-repo directory, so a single rg from the workspace
root silently misses nine of the ten repos. The real surface was mapped with
per-repo, per-family patterns. Full classification (remove / archive /
keep-historical / testnet-only-ok) is in INVENTORY.md in the
web3settle-monorepo PR.

The whole PR set

Repo	PR
smart-contracts	https://github.com/web3settle/smart-contracts/pull/13
merchant-sdk	#46
gateway-core-backend	https://github.com/web3settle/gateway-core-backend/pull/9
merchant-portal	https://github.com/web3settle/merchant-portal/pull/9
gateway-backoffice-portal	https://github.com/web3settle/gateway-backoffice-portal/pull/9
www-web3settle-com	https://github.com/web3settle/www-web3settle-com/pull/6
gateway-infra	https://github.com/web3settle/gateway-infra/pull/8
merchant-integration-demo	https://github.com/web3settle/merchant-integration-demo/pull/7
web3settle-monorepo	https://github.com/web3settle/web3settle-monorepo/pull/4

⛔ MERGE = RED — wait for Marcelo

Not merged, not approved, not deployed, not published. Nothing touched mainnet.
tron-canary.yml remains hard-off. No Dependabot PR was merged.

Read humanpending.md (in the web3settle-monorepo PR) before merging — in
particular HP-2, a pre-deploy database check, and HP-1, decision D4a.