Skip to content

docs(gaps): record the engine review as gaps 33–44 - #25

Merged
mo4islona merged 1 commit into
masterfrom
docs/fleet-readiness-gaps
Sep 5, 2026
Merged

docs(gaps): record the engine review as gaps 33–44#25
mo4islona merged 1 commit into
masterfrom
docs/fleet-readiness-gaps

Conversation

@mo4islona

@mo4islona mo4islona commented Sep 5, 2026

Copy link
Copy Markdown
Collaborator

What

spec/GAPS.md rewritten from a review of this engine against the reference.
make spec-check is clean; no other file changes.

  • Gaps 33–44 in the register's own format (invariant, severity, reproduction,
    first test). Three S1: the budget early-stop on block-sorted tables drops whole
    blocks and moves lastBlock past them, thread-count dependent (33); a filter on
    a column stored at a physical type outside the predicate's downcast matrix
    matches nothing (40); roll fields tolerate absent source columns (41). Gaps 40
    and 41 need a chunk of an older archiver vintage to reach, so their rating is
    conditional and the register says so. One S2: the catalogs lag the reference by
    thirteen fields, and transactionConfig cannot be expressed without struct
    member encodings (35). Four S3 and three S4, including the relation path
    materialising every target before the budget (36).
  • Deliberate-divergences table gains the rows the review found undocumented —
    malformed hex in a list, null filter values, the hyperliquidReplicaCmds
    alias predicates, shapes the reference rejects and this engine accepts. Two
    stale rows removed.
  • "Outside the engine" — what the suite actually proves, and the three matrix
    rows that overstate their evidence.

How it was checked

Both engines run side by side over the real EVM and Solana chunks, every fixture
chunk and ~70 synthetic chunk variants; ~330 request probes, ~700 response and
pagination runs, peak-memory measurements, and the full request/field surface
diffed against the reference's macros. Every number in the file was produced by a
run on this branch's tree; the scratch harnesses are not part of the change.

What this does not do

It records; it decides nothing. Whether malformed-hex and null-filter strictness
holds during the migration, and whether three matrix rows drop from C to P,
are left as open questions in the text.

Findings about the seams outside this engine — deployment, distribution and
routing — are tracked privately and are not in this file.

@mo4islona
mo4islona force-pushed the docs/fleet-readiness-gaps branch from 54ff575 to a1fac66 Compare September 5, 2026 11:54
@mo4islona mo4islona changed the title docs(gaps): record the fleet-readiness review as gaps 33–44 docs(gaps): record the engine review as gaps 33–44 Sep 5, 2026
@mo4islona
mo4islona force-pushed the docs/fleet-readiness-gaps branch 2 times, most recently from 54ff575 to 663735c Compare September 5, 2026 12:09
The register now has to say what a client would see, not only
where the engine and the spec disagree. The new entries come from
running both engines side by side over the real chunks, every
fixture chunk and ~70 synthetic chunk variants; the divergence
table gains the rows that review found undocumented.

Co-Authored-By: Claude Code <noreply@anthropic.com>
@mo4islona
mo4islona force-pushed the docs/fleet-readiness-gaps branch from 663735c to 032bf91 Compare September 5, 2026 12:59
@mo4islona
mo4islona merged commit 6df565a into master Sep 5, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant