fix(discovery): use actual slab size in memcmp fallback, not hardcoded V12_1 large

[6figpsolseeker]

Summary

Both memcmp fallback paths in discoverMarkets hardcoded maxAccounts: 4096 and dataSize: SLAB_TIERS.large.dataSize for all returned accounts. This caused detectSlabLayout to fail for any slab that wasn't V12_1 large, silently dropping valid markets with an "unrecognized layout" warning.

The root cause: memcmp queries used dataSlice (1940 bytes), making data.length always 1940. The fake dataSize hint then fed a V12_1-large size to detectSlabLayout, which only matches actual V12_1-large slabs.

Changes

• src/solana/discovery.ts:686-710 — removed dataSlice from the empty-result memcmp fallback; detect layout from actual data.length
• src/solana/discovery.ts:717-740 — same fix for the error-catch memcmp fallback

Both paths now use detectSlabLayout(data.length, data) to derive maxAccounts and dataSize from the actual slab data.

Trade-off

Higher bandwidth for the memcmp path (full slab data vs 1940-byte slice). Acceptable because this is already a last-resort fallback (tiers 1 and 2 both failed), and correct layout detection is worth the extra bytes.

Test plan

• npm run lint clean
• Full vitest run: same 14 pre-existing failures, zero new failures
• Recommended: add test covering memcmp fallback with non-large slab sizes

Summary by CodeRabbit

• Bug Fixes
  • Improved market discovery fallback logic when initial queries return no results. The system now fetches complete market data and dynamically calculates account sizing for better accuracy, replacing previously hardcoded size assumptions.

[coderabbitai]

📝 Walkthrough

Walkthrough

The fallback logic in discoverMarkets now fetches full slab accounts and dynamically detects their layout per account using detectSlabLayout, replacing hardcoded large-tier sizing. The resulting entries derive maxAccounts from the detected layout and set dataSize to the computed per-account size, applied consistently across both fallback paths.

Changes

Cohort / File(s)	Summary
Discovery Fallback Refactor src/solana/discovery.ts	Modified tiered fallback and memcmp fallback paths to fetch full slab accounts (no dataSlice), compute per-account actualSize from account.data.length, call detectSlabLayout(actualSize, data) for each account, and derive RawEntry objects using layout?.maxAccounts (default 4096) and dataSize=actualSize, removing hardcoded large-tier sizing.

Sequence Diagram

sequenceDiagram
    participant discoverMarkets
    participant RPC as RPC / getProgramAccounts
    participant detectLayout as detectSlabLayout
    participant results as RawEntry Results

    discoverMarkets->>RPC: Tiered queries with dataSlice
    RPC-->>discoverMarkets: Zero results
    
    discoverMarkets->>RPC: Fetch full slab accounts (no slice)
    RPC-->>discoverMarkets: Full account data

    loop For each account
        discoverMarkets->>discoverMarkets: Compute actualSize from account.data.length
        discoverMarkets->>detectLayout: detectSlabLayout(actualSize, data)
        detectLayout-->>discoverMarkets: Detected layout with maxAccounts
        discoverMarkets->>results: Build RawEntry with layout.maxAccounts & actualSize
    end
    
    results-->>discoverMarkets: Fallback discovery results

Loading

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~45 minutes

Possibly related PRs

• fix: add V1M2 slab layout for mainnet 312-byte account slabs #86 — Adds data-aware detectSlabLayout function and V1M2 layout structure with layout.maxAccounts field that this PR directly depends on for per-account detection.
• fix(slab): PERC-8469 — correct all V1M2 engine offsets from BPF offset table #125 — Updates slab layout logic (V1M2 engine offsets and size computations) that the per-account detectSlabLayout calls now rely on for accurate detection.
• fix: add mainnet medium tier slab size (257,512) — unblocks market discovery #29 — Introduces mainnet medium SLAB_TIERS entry that affects the slab size detection and layout derivation used in the modified fallback paths.

Poem

🐰 A rabbit hops through slabs so deep,
No more hardcoded tiers to keep,
Each account reveals its shape,
With layouts drawn from data's drape, 🔍
The fallback bounds now fit just right!

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main fix: using actual slab size in memcmp fallback instead of hardcoded large-tier values, which aligns with the core problem and solution described in the PR objectives.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

🧹 Nitpick comments (1)

src/solana/discovery.ts (1)

728-736: LGTM — consistent fix applied to the error-catch fallback path.

The same pattern is correctly applied here, ensuring both fallback paths behave identically.

Minor note: the mapping logic on lines 702-710 and 728-736 is now duplicated. Consider extracting to a small helper function for DRY, though this is optional given the limited scope.

,

♻️ Optional: Extract shared mapping logic

// Helper near the top of discoverMarkets or as a module-level function
function toRawEntries(
  accounts: Awaited<ReturnType<Connection["getProgramAccounts"]>>,
): RawEntry[] {
  return accounts.map(e => {
    const actualSize = e.account.data.length;
    const layout = detectSlabLayout(actualSize, new Uint8Array(e.account.data));
    return {
      ...e,
      maxAccounts: layout?.maxAccounts ?? 4096,
      dataSize: actualSize,
    };
  });
}

Then both fallback paths become:

rawAccounts = toRawEntries(fallback);

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/solana/discovery.ts` around lines 728 - 736, The mapping that converts
getProgramAccounts results into RawEntry objects is duplicated (lines mapping to
rawAccounts in discoverMarkets); extract that logic into a helper (e.g.,
toRawEntries) that accepts the accounts array, computes actualSize, calls
detectSlabLayout(new Uint8Array(...)), sets maxAccounts = layout?.maxAccounts ??
4096 and dataSize = actualSize, and returns RawEntry[]; then replace both inline
map usages with rawAccounts = toRawEntries(fallback) (and the other occurrence)
to remove duplication while preserving behavior.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@src/solana/discovery.ts`:
- Around line 728-736: The mapping that converts getProgramAccounts results into
RawEntry objects is duplicated (lines mapping to rawAccounts in
discoverMarkets); extract that logic into a helper (e.g., toRawEntries) that
accepts the accounts array, computes actualSize, calls detectSlabLayout(new
Uint8Array(...)), sets maxAccounts = layout?.maxAccounts ?? 4096 and dataSize =
actualSize, and returns RawEntry[]; then replace both inline map usages with
rawAccounts = toRawEntries(fallback) (and the other occurrence) to remove
duplication while preserving behavior.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: c44929be-488a-4dd1-81d2-c6f760a9ac91

📥 Commits

Reviewing files that changed from the base of the PR and between 2f80d79 and fda3e6a.

📒 Files selected for processing (1)

• src/solana/discovery.ts