docs: add complete public RPC reference - #85
Conversation
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
MegaETH Public RPC Reference — PR Review
Verdict: APPROVE with revisions requested — 1 Major finding, 3 Minor findings, 1 Nit (pre-mortem, unverified). Nothing blocks merging technically, but the Major finding is a silent regression readers will hit before a follow-up ships.
Note: this review is body-only. All findings are consolidated below with file:line references; each was verified against the head tree of this PR.
What is good
- Comprehensive coverage: a dedicated reference page per method with consistent frontmatter and
Parameters/Returns/Errors/Examplestructure, plus per-errorCode | Cause | Fixtables. A clear step up from the prior overview-only doc. - Availability probes are dated (
2025-11-11in the intro ofdocs/dev/read/overview.md) — the right freshness signal. - OP-Stack aliasing is called out for
mega_getWithdrawalProof/eth_getWithdrawalProof— a good pattern that should be repeated (see Minor #2). - Uncle-block behavior on
eth_getUncle*is honestly documented (returns null / always 0 on this non-uncle chain) instead of silently omitted. - The Cache section on
eth_callis candid about staleness and eventual consistency instead of pretending strong read guarantees. - The new index
docs/dev/read/rpc/README.mdgives a clear method → page map and is registered indocs/SUMMARY.md.
Findings
[Major] (pre-mortem, confirmed) docs/dev/read/overview.md:74 — Six filter methods present in the base availability table were silently dropped from the new table: eth_newFilter, eth_newBlockFilter, eth_newPendingTransactionFilter, eth_uninstallFilter, eth_getFilterChanges, eth_getFilterLogs. eth_mining was preserved and reclassified to Unavailable, so the omission of the filter set stands out.
The PR description commits to "list gateway-recognized but unavailable methods separately so the public docs do not promise disabled APIs" — but a reader of the base doc who was polling with eth_newFilter + eth_getFilterChanges now cannot tell whether the method was removed from the gateway, whether it still works, or whether the docs are stale. This is a silent regression relative to the prior page.
Please either re-add these six methods as Unavailable rows with the -32601 note (matching the eth_callAfter / eth_getLogsWithCursor pattern already used in this PR), or explicitly note in the page intro that filter-based polling is no longer registered on the public gateway. Verification: probe each of the six method names against https://mainnet.megaeth.com/rpc and record the response.
[Major] docs/dev/send-tx/debugging.md:19 — Not modified by this PR, but this PRs new docs/dev/read/overview.md:38 marks debug_traceCall as Unavailable with -32601, while debugging.md:19 still states debug methods including debug_traceCall are supported on the public MegaETH RPC endpoint. Readers who follow the debugging guide will hit -32601 from the public gateway. Please reconcile — either update debugging.md in this PR to note debug_traceCall is unavailable on the public gateway (mega-evme or a managed provider must be used), or remove the Unavailable classification from overview.md.
[Minor] (pre-mortem, confirmed) docs/dev/read/rpc/eth_sendRawTransactionSync.md and docs/dev/read/rpc/realtime_sendRawTransaction.md — Both methods submit a signed tx, wait for execution, and return a receipt inline; both document the same 5s default and 3000ms public-gateway cap. Yet neither page references the other, and there is no explanation of which name a client should prefer or whether they are aliases. Please add a one-line cross-reference from each page to the other (following the mega_getWithdrawalProof / eth_getWithdrawalProof aliasing pattern already used in this PR) and clarify the relationship — same underlying handler, or two distinct entry points with subtle differences?
[Minor] docs/dev/read/rpc/mega_outputAtBlock.md:92 — The example response mixes what looks like a real captured payload (outputRoot, blockRef.hash, withdrawalStorageRoot, stateRoot all look like real hashes) with obviously synthetic placeholders (parentHash: 0xa1b2c3d4e5f60718293a4b5c6d7e8f900112233... and l1origin.hash: 0x0011223344556677889900aabbccddeeff...). Readers copying this to build test fixtures may not notice which fields are illustrative vs. captured. Either replace all fields with a single consistent captured response for block 0x100, or make every hash obviously synthetic (e.g., all-zeros with a trailing tag).
[Minor] docs/dev/read/rpc/eth_sendRawTransactionSync.md:100 — The example response shows blockHash: 0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff — an all-1s sentinel that is not standard Ethereum output. If this is a MegaETH-specific sentinel (e.g., "executed in a mini-block that has not yet been committed to a full block hash"), it deserves a one-line note in the Returns table so downstream indexers do not treat it as a real hash. If it is just a placeholder chosen at random, please replace it with a plausible-looking hash to avoid confusion.
Pre-mortem (unverified)
[Nit] lychee.toml — include_fragments changed from "full" to true. This looks like a config-syntax normalization, but if lychee treats true as a stricter fragment mode than the previous string value, the next CI run may start flagging fragment links (e.g., page.md anchor references) that previously passed. Please run lychee locally against the current tree with the new value and confirm no previously-passing links now fail; if any do, either fix them in this PR or roll back to the string form.
Verification receipts
- Confirmed the six dropped filter methods were present in base
docs/dev/read/overview.mdand are absent from the head version. - Confirmed
realtime_sendRawTransaction.mdand neweth_sendRawTransactionSync.mdboth document the same 5s default + 3000ms public-gateway cap without cross-referencing each other. - Confirmed
docs/dev/send-tx/debugging.md:19was not touched by this PR and still promises debug methods on the public endpoint.
There was a problem hiding this comment.
🧭 Re-review update — 3 new findings (pre-mortem)
Reviewed against head SHA b363af0 — unchanged since the prior review. No code has been updated, so every prior finding still stands (recapped below). Three new findings from a fresh pre-mortem pass are added as body-only bullets because the standard pending-review-inline-comment API path was not reachable from this environment on retry — the findings are file:line anchored so they remain actionable.
New pre-mortem findings (confirmed):
-
[Major]
docs/dev/read/rpc/debug_getRawHeader.md:79-80(and 9 sibling pages:debug_traceBlockByHash.md,debug_traceBlockByNumber.md,debug_traceTransaction.md,eth_callMany.md,eth_getUncleByBlockHashAndIndex.md,eth_getUncleByBlockNumberAndIndex.md,eth_getUncleCountByBlockHash.md,eth_getUncleCountByBlockNumber.md,eth_protocolVersion.md) — every "Sources" section links togithub.com/megaeth-labs/mega-rethandgithub.com/megaeth-labs/mega-rpc, which both return HTTP 404. Neither repo is in themegaeth-labspublic repo list (which containsreth,mega-evm,stateless-validator, etc.). Readers who click a source link to verify a claim will get a 404, inverting the trust value of the newly added citation apparatus. CI does not catch this becauselychee.tomlhasoffline = true. Fix: point at the actual public repos (e.g.,megaeth-labs/rethfor the fork), drop the Sources section on pages whose source lives in a private repo, or run lychee non-offline in a scheduled workflow. -
[Minor]
docs/dev/read/rpc/eth_getWithdrawalProof.md:14— Says theaddressparameter must be0x4200000000000000000000000000000000000016, but the aliased method (docs/dev/read/rpc/mega_getWithdrawalProof.md:11-14) describes the same parameter as "the contract address to prove" with that L2ToL1 address only as a recommendation. Line 8 of the same file promises both names "route to the same node implementation" — so one of the two descriptions is wrong for the shared handler. Reconcile: either both pages fix the address to0x4200…0016, or both describe it as an arbitrary contract with the L2ToL1 address as a recommendation. -
[Minor]
docs/dev/read/rpc/README.md:10— Directs readers to the availability table "for … methods offered only by managed providers", but the table inoverview.mdno longer contains any managed-only rows:debug_getRawBlock,debug_getRawReceipts,debug_getRawTransaction,debug_replayBlock,debug_traceBlock,debug_traceCallMany,trace_callMany,trace_get,trace_rawTransaction,trace_replayBlockTransactions, andtrace_replayTransactionwere all deleted along with the "Managed only" classification. A reader looking up (say)debug_getRawBlockwill find no entry and conclude MegaETH does not support it anywhere. Either restore a "Managed providers" subtable inoverview.md, or drop the managed-providers phrase from this pointer.
Still open from the prior review (head SHA unchanged):
- [Major]
docs/dev/read/overview.md:74(and surrounding table) — Six filter methods (eth_newFilter,eth_newBlockFilter,eth_newPendingTransactionFilter,eth_uninstallFilter,eth_getFilterChanges,eth_getFilterLogs) that wereAvailablein the base table are silently absent from the new table. Either re-add them asUnavailablerows with-32601, or explicitly note that filter-based polling is no longer registered on the public gateway. - [Major]
docs/dev/send-tx/debugging.md:19still promisesdebug_traceCallon the public endpoint; this PR marks itUnavailableatdocs/dev/read/overview.md:38. Reconcile in this PR to avoid a silent doc-to-doc contradiction. - [Minor]
docs/dev/read/rpc/eth_sendRawTransactionSync.mdanddocs/dev/read/rpc/realtime_sendRawTransaction.mdcover the same synchronous submit-and-receipt flow but never cross-reference each other or explain the relationship (same handler? separate entry points?). Add a one-line cross-reference from each page to the other. - [Minor]
docs/dev/read/rpc/mega_outputAtBlock.md:92,95mixes real-looking captured hashes with obviously synthetic placeholders in a single example (parentHashandl1origin.hashare pattern-fillers next to real-looking values elsewhere). Make the example uniformly real or uniformly synthetic. - [Minor]
docs/dev/read/rpc/eth_sendRawTransactionSync.md:100showsblockHash: 0xffff…ffffsentinel without explanation. If it is a MegaETH pre-full-block sentinel, add a one-line note to the Returns table; otherwise replace it with a plausible-looking hash.
Pre-mortem (unverified):
docs/dev/read/overview.md:43,61,63-64— Four methods (eth_callAfter,eth_getLogsWithCursor,eth_getTransactionByBlockHashAndIndex,eth_getTransactionByBlockNumberAndIndex) flip fromAvailable/Managed onlyin the base table toUnavailablehere. The PR description mentions probing the public endpoint but no raw probe artifact is linked in the diff. Please attach the curl transcript per contested row (or link one in the PR body) — two of these are standard library methods, and a wrong classification will cascade into third-party library fallback paths.
Pre-mortem: 3 candidates confirmed and reported above; 1 candidate ranked plausible-unverified is included as a verification request; 1 duplicate of the prior review (0xffff…ffff sentinel) suppressed; 1 rejected.
Delivery note: this round was intended to use the pending-review + inline-comments flow (see PR review contract). The POST /pulls/85/reviews/<review_id>/comments endpoint returned HTTP 404 from this environment on retry, so the pending review was deleted and all findings were consolidated into this single review body with explicit path:line anchors. Every finding is still actionable at those line references.
|
Addressed all actionable findings in
The PR description now includes the requested raw public-endpoint probe outcomes. The final inventory comparison is exact: 51 expected public methods and 51 reference pages. |
There was a problem hiding this comment.
🧭 Re-review update — all prior findings addressed; 2 new pre-mortem findings
- Reviewed head SHA
34b4a74against the prior review atb363af0. Every finding from the two previous rounds has been addressed in the new commit (docs: address RPC reference review feedback). - Nice work on the fixes: the six missing filter methods are now in the availability table,
debugging.mdandoverview.mdagree ondebug_traceCall, the brokenmega-reth/mega-rpcsource links have been rewritten to internal-repo citations, theblockHashall-ffsentinel is now explained on both sync-submit pages,mega_outputAtBlockuses a uniformly synthetic example, and the two withdrawal-proof pages agree on the parameter semantics. - New findings from this round: 2, both body-only because their anchor lines are outside the RIGHT-side of the PR diff.
Body-only findings:
-
[Major] (pre-mortem, confirmed)
docs/dev/send-tx/overview.md:122— Still reads "Times out after 10 seconds if the transaction has not been executed." forrealtime_sendRawTransaction. This PR corrected the same figure indocs/dev/read/realtime-api.mdand the newdocs/dev/read/rpc/realtime_sendRawTransaction.md(node default 5s, public gateway caps explicit timeouts at 3,000ms). Leaving the 10s figure on the Send Transaction landing page is a silent contradiction — a developer following that page will size their client wait budget to 10s and treat every capped submission as a false failure. Please reconcile in this PR, since the PR description explicitly commits to "correct stale … realtime timeout … guidance".Suggested wording:
Returns the full transaction receipt directly. The node waits up to 5 seconds by default; the public gateway caps an explicit `timeout` at 3,000 milliseconds and returns `-32000 realtime transaction expired` if the deadline elapses first. See [`realtime_sendRawTransaction`](../read/rpc/realtime_sendRawTransaction.md) for the full reference. -
[Minor] (pre-mortem, confirmed)
docs/dev/read/overview.md:108anddocs/dev/read/overview.md:117— The Rate Limiting section still listseth_callAfterin the Simple category row and adds a bullet explaining its per-call budget, even though this PR reclassifieseth_callAfterasUnavailableat line 43 (the wholeeth_callAfter.mdreference page was deleted). The two mentions are stale and contradict the availability table in the same file. Dropeth_callAfterfrom the Simple-row description on line 108 and delete the line-117 bullet; the section already covers "all other read methods not listed in another category."
Pre-mortem (unverified):
docs/dev/send-tx/debugging.md:15now tells developers to "usemega-evmefor local call simulation or check whether a managed RPC provider exposes that method" in place ofdebug_traceCall. Please verify thatmega-evme(or a specific managed provider) actually documents a working call-simulation path against Mainnet state at head, so users who follow the pointer land on a runnable substitute rather than an unsupported use case.
Pre-mortem: 2 confirmed and reported above; 1 plausible-unverified included as a verification request; 5 candidates rejected or suppressed.
|
Follow-up review addressed in
The reported |
There was a problem hiding this comment.
🧭 Re-review update — 1 new finding
Reviewed head c645a25d against the prior review at 34b4a74. The send-tx/overview.md 10s→5s reconciliation and the in-page mega-evme link both address the prior findings cleanly.
Inline anchoring on the pending review returned HTTP 404 from this environment (same failure mode as review round 2 documented in-thread), so both findings below are body-only with explicit path:line anchors. They remain actionable at those references.
Findings:
-
[Major] (pre-mortem, confirmed)
docs/dev/read/rpc/eth_getBalance.md:34-35— the Errors table cites-32001and4444, and line 37 links toerror-codes.mdfor followup. Butdocs/dev/read/rpc/error-codes.mdat head only documents-32700,-32099,-32601,-32019,-32020,-32021,-32022,-32000— neither cited code appears. Same dead-end pattern ineth_getBlockByHash.md:64(4444),eth_getBlockByNumber.md:64(4444),eth_getBlockReceipts.md:70(4444),eth_getCode.md:33-34(-32001,4444),eth_getStorageAt.md:39-40(-32001,4444),eth_getTransactionCount.md:34-35(-32001,4444), anddebug_getHistoryTransactionCount.md:27(-32001). Fix: add rows for-32001(block selector cannot be resolved) and4444(historical state not available on this endpoint) toerror-codes.md, or drop theSee also [Error reference]pointer from pages that cite unlisted codes. -
[Major] (merge-conflict alert)
docs/dev/read/overview.md— PR 85 head has the simple CU-based Rate Limiting section (noeth_callAfter). Since PR 85 branched,main(via PR #83) replaced that section with the older four-tier table that still listseth_callAfterin the Simple category and adds a per-method note about its 60M compute-gas cap. Because PR 85 does not touch those lines, the 3-way merge preserves main version, and the deployed page will contradict itself (eth_callAftermarkedUnavailablein the availability table at:43, described as an active Simple-category method in Rate Limiting at merged:108and:117). Please rebase against main and reconcile — either drop theeth_callAfterreferences from the older table or replace that section with the simple version.
Pre-mortem (unverified):
- The PR deletes
docs/dev/read/rpc/eth_callAfter.md(−123 lines) anddocs/dev/read/rpc/eth_getLogsWithCursor.md(−128 lines) without adding a.gitbook.yamlredirect. If either slug has meaningful external inbound links (the PR description mentions the pages were previously drafted ondeveloper-docs), the URLs will 404 after deploy. Confirm the inbound traffic to those slugs before merge, or add aredirects:block pointing them atdev/read/overview.md#available-methods.
Pre-mortem: 1 confirmed (error-codes.md gap, above), 1 plausible-unverified included as verification request, 1 rejected at head SHA but preserved as merge-conflict alert (Rate Limiting eth_callAfter — refuted against c645a25d but reappears after 3-way merge with main), 2 low-confidence hypotheses suppressed (GitBook anchor slugs, cast rpc snippet in the managed-provider tab).
…c/complete-rpc-reference
|
Addressed the latest review in
Validation after the merge:
|
There was a problem hiding this comment.
⚠️ 2 findings
Since the last Claude bot review, the branch has cleanly addressed every prior-round finding: -32001 and 4444 are now defined in docs/dev/read/rpc/error-codes.md:12-13; the RPC index no longer calls the disabled methods "managed-only" (docs/dev/read/rpc/README.md:10); redirects for eth_callAfter / eth_getLogsWithCursor were added in .gitbook.yaml; and the debug_traceCall availability caveat is stated in docs/dev/send-tx/debugging.md:15. The 51-method total in docs/dev/read/rpc/README.md:7 matches docs/SUMMARY.md.
Two new issues found in this pass are folded into the summary below (inline comment posting was blocked by tooling constraints in this environment).
Major
docs/dev/read/overview.md:114— The read-rate-limit exemption bullet lists onlyeth_sendRawTransactionandrealtime_sendRawTransaction, but this PR also introduceseth_sendRawTransactionSyncas a first-class transaction-submission method (seedocs/dev/read/rpc/eth_sendRawTransactionSync.md:8, which states the public gateway routes it through the same synchronous submission handler asrealtime_sendRawTransaction). Readers will reasonably assumeeth_sendRawTransactionSyncfalls under the read rate-limit tier documented above the bullet. Addeth_sendRawTransactionSyncto the exemption list, and mirror the addition in the Request Body Limits table atdocs/dev/read/overview.md:125(which currently lists only the other two submission methods at 2.5 MiB).
Minor
docs/dev/read/rpc/eth_uninstallFilter.md:78— The "Source" footer exposes an internal-repo path:Code: mega-reth@0264d0821a8fe14ac6c7f710e9452edef7407b3f, crates/rpc/rpc/src/eth/filter.rs (internal repository). The public docs should not leak private commit SHAs or file paths — they are not verifiable by external readers and hint at implementation details that may change. Either drop the pointer, or link to a public equivalent (mega-evmetc.) if one exists. Worth a repo-wide check for any other pages in the same batch that carry an(internal repository)footer.
Pre-mortem (unverified)
.gitbook.yaml— The addedredirects:block uses source keys without a leading slash or.mdsuffix (e.g.,dev/read/rpc/eth_callAfter: dev/read/overview.md#available-methods). Worth confirming against GitBook's current redirect schema that source paths are matched as URL paths (no.md) while destinations are file paths (with.md) — the asymmetry is easy to get wrong and only surfaces after deploy.
|
Addressed both findings in
The GitBook redirect shape is also confirmed against the current official GitBook content-configuration documentation: redirect paths are relative to |
There was a problem hiding this comment.
✅ Clean re-review
- Reviewed the 12-file incremental diff since the prior review at
048439366(1 new commit:docs: complete public RPC policy guidance). - Both findings from the previous round are cleanly addressed:
eth_sendRawTransactionSyncis now listed in the read rate-limit exemption bullet (docs/dev/read/overview.md:114) and in the Request Body Limits table (docs/dev/read/overview.md:125), matching its documented parity withrealtime_sendRawTransaction.- Every
(internal repository)source footer that leaked privatemega-reth/mega-rpcSHAs and paths has been removed from all 11 pages that carried them; no other pages indocs/still reference them.
- The
.gitbook.yamlredirects foreth_callAfterandeth_getLogsWithCursorland on the availability table, which does list both asUnavailable, so the deprecation is discoverable. - No new actionable findings.
- Still open from earlier reviews: none.
Pre-mortem (unverified):
docs/dev/read/rpc/error-codes.md:13— The new4444/pruned history unavailablerow (also cited ineth_getBlockByHash.md:64andeth_getBlockByNumber.md:64) is not covered by the PR probe table. Worth confirming with a probe against a pre-retention / beyond-history block hash that the public gateway actually emits numeric code4444(not-32001or-32000); if the emitted code differs, integrator branches keyed on4444will silently never fire.
Pre-mortem: no verifiable high-impact failure path found in the incremental scope.
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
❓ Review complete — 1 open question(s)
Reviewed head 78f47e8b.
Reviewed the RPC reference migration: 47 new method pages, availability-table restructure, redirects for two removed pages, and related overview/FAQ/gas-estimation edits.
Open questions:
❓ Open question · Medium confidence
- Does GitBook's
.gitbook.yamlredirects:schema accept a target of the formpath/to/page.md#anchor, or does it require a bare page URL (no.md, no fragment)? If the current form is not the documented syntax,dev/read/rpc/eth_callAfteranddev/read/rpc/eth_getLogsWithCursorwill 404 after publish instead of redirecting to the availability table. - Why it matters: These are the only preservation path for inbound links to the two deleted pages (SDKs, blog posts, search-engine cached results). If the redirects are silently dropped at build time, users hit hard 404s on URLs that worked before the merge, and CI would not catch it (lychee only checks in-repo links).
- How to verify: After deploy, GET
https://docs.megaeth.com/dev/read/rpc/eth_callAfterandhttps://docs.megaeth.com/dev/read/rpc/eth_getLogsWithCursorand confirm the response is a 301/302 todev/read/overview#available-methods(or the fragment-less equivalent). If either 404s, change targets to bare page URLs without.md, e.g.dev/read/rpc/eth_callAfter: dev/read/overview, and confirm the GitBook build log shows the redirects registered.
Claude review statusLast reviewed: head New this round: 2 · Resolved this round: 1 · Open questions: 2 |
| Sending `Cache-Control: no-store` or `no-cache` as a _request_ header does not bypass the internal cache — request cache directives address intermediary caches, not the origin's own caching. | ||
| The gateway does not provide a request option that bypasses its internal cache. | ||
|
|
||
| ## Method Reference |
There was a problem hiding this comment.
[Nit] Empty ## Method Reference heading creates orphaned section in RPC index
Readers and GitBook's table-of-contents see a dangling section that contains no prose. Any anchor link to #method-reference lands on an empty heading, not a useful entry point.
Suggested fix: Remove line 135 (## Method Reference) and its trailing blank line (line 136). The subsections (## State and simulation, ## Blocks, transactions, and receipts, etc.) already self-organize the listing without a wrapper heading.
There was a problem hiding this comment.
0 blocking · 0 should-fix · 2 suggestion(s) · 2 open question(s)
Reviewed head f157be65.
Details are attached inline.
Open questions:
❓ Open question · Low confidence
- Should
docs/dev/rpc/operations-and-limits.mdname the four methods eligible for the gateway's internal cache (eth_getBlockByNumber,eth_getBlockReceipts,eth_getHeaderByNumber,web3_clientVersion) and the immutable-tag rule, the way the deleteddocs/dev/read/overview.mddid? - Why it matters: The new page describes gateway caching only in general terms. The removed page listed the exact methods and stated that only requests for an explicit historical block number, block hash, or
earliesttag are cache-eligible. If that concreteness matters to users (debugging stale-looking reads, understandingX-Workers-Cache-Status), the abstraction is a regression; if the list was expected to churn, the abstraction is intentional. - How to verify: Confirm with the RPC gateway owners whether the four-method allowlist and the immutable-tag rule are still stable enough to publish. If yes, add them back to
operations-and-limits.md#gateway-caching.
❓ Open question · Medium confidence
- In
docs/dev/rpc/reference/eth_getTransactionReceipt.md, the success example (lines 110–127) omitslogs,logsBloom, andcumulativeGasUsed, which the Result schema (lines 59–63) documents as present on every receipt. Should the example be regenerated with those fields, or should the schema note them as sometimes omitted by the gateway? - Why it matters: The mismatch pushes readers to write parsers that tolerate missing fields where the spec says they are required, or the other way around — depending on which side they trust. Either direction is a wrong impression from a reference page whose purpose is to pin the exact wire contract.
- How to verify: Re-run the documented probe (
eth_getTransactionReceiptfor0xf3473347…8ed6againsthttps://mainnet.megaeth.com/rpc) and either update the example to include the missing fields or add an 'abridged for readability' note (likemega_outputAtBlock.md:114uses) that points readers to the full schema.
| | `trace_transaction` | Unavailable | The public endpoint returns `-32601`. | | ||
| | `web3_clientVersion` | Available | | | ||
|
|
||
| ## Method Reference |
There was a problem hiding this comment.
[Nit] Empty ## Method Reference heading regenerated in the new reference index
GitBook's page TOC and any anchor link to #method-reference land on an empty section. This is the same defect the prior review flagged on the old file path; the refactor relocated it rather than removing it.
Suggested fix: Remove line 86 (## Method Reference) and its trailing blank line. The subsequent H2s (## State and simulation, ## Blocks, transactions, and receipts, …) already organize the listing without a wrapper heading.
| ```json | ||
| { | ||
| "jsonrpc": "2.0", | ||
| "id": 1, | ||
| "result": { |
There was a problem hiding this comment.
[Minor] realtime_sendRawTransaction example receipt omits documented blockHash field
A client author reading only this example will treat blockHash (and other fields) as optional/absent, contradicting the documented schema and the parallel eth_sendRawTransactionSync example. This is exactly the kind of divergence the PR is trying to eliminate between the two aliased methods.
Suggested fix: Regenerate the example from a real probe so it matches the schema (include blockHash, transactionIndex, cumulativeGasUsed, effectiveGasPrice, L1 fee fields, etc., or explicitly note the example is abridged and the placeholder rule from line 40 applies). Keeping the two example receipts structurally identical would make the 'same handler' claim self-evident.
1 participant
Summary
developer-docsJSON-RPC→RPC Reference→ individual method pagesDocumentation structure
docs/dev/rpc/reference/.Notable corrections
eth_subscribedocuments all six supported subscription types:newHeads,logs,newPendingTransactions,syncing,miniBlocks, andstateChanges.stateChangesdocuments that an omitted or empty address list subscribes to all changed accounts.newHeadsandminiBlocksinclude the MegaETH-specific metadata and payload fields implemented by the node.-32019and-32020claims.200.eth_getLogsno longer claims a fixed 100-block public gateway cap on the indexed path.eth_feeHistory,eth_accounts, andeth_maxPriorityFeePerGasdocument their gateway-synthesized results.eth_calldocuments its 60,000,000 compute-gas override, whileeth_estimateGasdocuments its separate CPU-time-limited path.eth_sendRawTransactionSyncis identified as a gateway compatibility alias rather than a native node method.Validation
mise run check-links: 938 links checked, 0 errorsmise run check-lint: 108 Markdown files, 0 issuesmise run check-format: all Markdown files match the pinned Prettier versiongit diff --checkPublic endpoint verification highlights
eth_feeHistoryeth_accounts[]eth_maxPriorityFeePerGas0x0eth_getBlockByNumberwithpendingnullpendingeth_getWithdrawalProofmega_outputAtBlockversion-32601eth_subscribetypesstateChangeswithout an address filter