feat: P1.5 engine diag counters — submitted/terminal/overflowRej on EGRESS-DIAG#42
Merged
Merged
Conversation
…GRESS-DIAG Adds the correctness counters the P1 gate harness reads (match#32), modeled on the existing zero-alloc egress diag pattern: - ordersSubmitted: CreateOrder commands accepted off ingress (plain long in SbeDemuxer; written and read on the cluster service thread only) - ordersTerminalized: terminal OrderStatus events (FILLED/CANCELLED/REJECTED) counted in MatchEventPublisher BEFORE the Disruptor, so egress shedding cannot hide a terminalization (plain long, service thread only) - overflowRejects: pre-plumbed AtomicLong + new OrderRejectReason.OVERFLOW=5. Stays 0 until P1.1 (match#30) makes FixedPoint.multiply throw on true overflow and routes the reject here; the harness records the 0 baseline. All three appear on the periodic EGRESS-DIAG leader-health line and the role-transition dump (submitted=/terminal=/overflowRej=). No allocation added to the hot path; strings stay confined to the 5s diag timer. Part of match#32 (P1.5). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
emrebulutlar
added a commit
that referenced
this pull request
Jul 3, 2026
…LOW rejects at admission (#52) FixedPoint.multiply picked its path by the FIRST operand's magnitude, so multiply(smallQty, largePrice) wrapped silently (any qty-first notional over ~$92B raw product), products in [2^63, 2^64) came out NEGATIVE (the ~$922- $1844 band that spuriously NOTIONAL_TOO_SMALL-rejects ~3% of typical orders via the OMS risk path), and the 128-bit branch divided by an APPROXIMATION of 2^64/SCALE. divide's wide path wrapped on both of its terms. Both are now exact with truncation toward zero, selecting the wide path by whether the PRODUCT fits 64 bits (operand order can never matter again), and the wide path is an exact unsigned 128/64 division (Hacker's Delight divlu). When the true result does not fit a signed long they throw FixedPoint.OverflowException (stackless: cheap under hostile input) — wrong money is never returned. Engine routing: LIMIT / LIMIT_MAKER / UPDATE admission validates the notional alongside validateLimitPrice and rejects with OrderRejectReason.OVERFLOW (counted on the EGRESS-DIAG line as overflowRej, replacing the #42 pre-plumbed AtomicLong). Since no order with an unrepresentable notional can enter the book, every downstream cost product is bounded by an admitted notional and cannot overflow mid-match; the market-buy budget divisions (the only remaining throw sites) stop matching deterministically instead. Property tests: BigInteger oracle over 1M+ samples — edge grid, financial ranges (prices >= $10M, qty to book capacity), products concentrated in 2^62..2^66 (the band and the old approximate path), uniform random longs, both operand orders every time, multiply/divide round-trip. Determinism corpus: unchanged (no golden shifts). Fixes #30. Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Part of #32 (P1.5). The correctness counters the P1 gate harness reads, modeled on the existing zero-alloc egress diag pattern:
SbeDemuxer; written and read on the cluster service thread only)MatchEventPublisherBEFORE the Disruptor, so egress shedding cannot hide a terminalizationOrderRejectReason.OVERFLOW=5. Stays 0 until P1.1 ([P1.1] FixedPoint.multiply: product-overflow-safe + throw; route match notional sites #30) makesFixedPoint.multiplythrow on true overflow and routes the reject here; the harness records the 0 baselineAll three appear on the periodic
EGRESS-DIAG leader-health:line and the role-transition dump. No allocation added to the hot path; strings stay confined to the 5s diag timer.mvn clean verifygreen.🤖 Generated with Claude Code