feat(infer): per-module resolve memo (Option B) — within-module type-grounding memo by briansrls · Pull Request #6779 · gunb-ai/gunbc

briansrls · 2026-07-16T22:41:36Z

Summary

Transient per-module memo at resolve_node_bounded grain (Option B). Replaces dead Option-2 persistent census pre-resolve (OOM at corpus scale).

Content-identity keys (dag_node_surface_fingerprint + depth + masked); ABA-safe pinned fp cache
Scrutinee parallel memo; hit/miss/bypass counters; re-entrancy-safe probe-then-compute-then-store
Composes with clever-seal cross-module memo (Lever 1: floor compile-clean receipt through the shared index — one typecheck universe per run #6783); no cli_run collision

Measured main-tree benefit (clever-seal whole-tree RNB attribution, 1090 modules)

Typecheck fold 345.7s; resolve_node_bounded 43.1s = 12.5% of fold (733k calls)
Main is not resolve-bound in aggregate — heterogeneous by module
Primary customer: host_effect_realize — 70% rnb (20.0s of 28.7s module time, 108,941 calls)
Other modules low: ci_spec 1.2%, fleet_converge_emit 0.1%, srv3 receipt 0.7%
Honest standalone case: ~up to ~40s (~12%) typecheck win, concentrated in host_effect_realize, plus strip-lane bounded-call hygiene

Not the #6640 unblocker (root: peel_alias_once_for_field_access infinite loop — clever-seal peel guard). Not an 82%/434s main-wall claim.

Test plan

cargo test -p v1-compiler-tests per_module_resolve_memo_equivalence
Operator release (b): memo verified; peel loop diagnosed as root cause

Head

session/census-pre-resolve @ 9215b0a712 (rebased on still-owl-689)

cursor · 2026-07-17T01:14:17Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

gunbai-bot · 2026-07-17T01:15:38Z

Addressed review 39138 (env-var test flakiness):

Finding valid. set_var/remove_var on GUNBC_PER_MODULE_RESOLVE_MEMO was process-global and unsafe under parallel cargo test.

Fix (aafefa03d8): thread-local PerModuleResolveMemoEnabledOverride + PerModuleResolveMemoEnabledGuard (#[doc(hidden)]). Receipt tests use PerModuleResolveMemoEnabledGuard::force(false) for the cold path; no env mutation. Production still honors GUNBC_PER_MODULE_RESOLVE_MEMO when override is None.

Style nit (generic memo helper): agreed as follow-up if seed lingers; not blocking given explicit DISSOLVE-ON to v2.std.materialize.

— sent from loyal-crab-450

gunbai-bot · 2026-07-17T01:31:21Z

Review 39142 (addressed in latest push):

module_name excluded from BoundedMemoKey — documented invariant at BoundedMemoKey: within one install window every resolve_node_bounded call uses the same enclosing resolved_module_name; module_name labels diagnostics only and does not change the grounded resolved form for identical (env, node, depth, masked).
Panic-safe scope — per_module_resolve_memo_install now flushes+log any leaked scope instead of assert!; PerModuleResolveMemoScopeGuard wraps infer+pop in typecheck_module and flushes on Drop.

Review 39143 — no findings; thanks for APPROVE.

— sent from loyal-crab-450

gunbai-bot · 2026-07-17T01:40:13Z

Release (b) verdict (operator, quiet-gull): memo layer is correct — content-key captures all distinct work (~1787 misses, then ~100% hits); re-entrancy fix holds. Does not unblock #6640: whole-tree still stalls in test.claim.bmc_bootstrap_provision_witness due to caller-level superlinear resolve_node invocation (expand_type_for_field_access → peel_alias_once_for_field_access under infer_expr), not per-call resolution cost.

Hold: do not merge as the #6640 fix. Standalone landing TBD after clever-seal re-prices main-434s gate. Receipt (a) + composite test retained.

— sent from loyal-crab-450

Within-module transient memo at resolve_node_bounded grain catches infer_items and populate_output_provenance re-grounding without the Option-2 persistent census roots that OOM'd. Content-identity keys (dag_node_surface_fingerprint + depth + masked); scrutinee parallel memo on resolve_scrutinee_type_node entry. Freed at module exit. Receipt (a): per_module_resolve_memo_equivalence_test. Co-authored-by: Cursor <cursoragent@cursor.com>

Makes release (b) diagnostic: bypass counts env-ptr guardrail skips separately from fingerprint misses. Restores ptr->fp hint cache (amortize hashing; content String remains map equality key). Co-authored-by: Cursor <cursoragent@cursor.com>

Retain (Rc<Node>, fp) in node_surface_fp_cache so transient wrapper addrs cannot be reused with stale fingerprints during the module scope. Co-authored-by: Cursor <cursoragent@cursor.com>

Probe-then-compute-then-store pattern lets resolve_node_bounded recurse through the memo wrapper without panicking. Strengthen receipt (a) with composite Optional<ProbeCurrency> resolve under installed memo scope. Co-authored-by: Cursor <cursoragent@cursor.com>

Replace process-global GUNBC_PER_MODULE_RESOLVE_MEMO set_var/remove_var with PerModuleResolveMemoEnabledGuard so parallel cargo test cannot flake the byte-identity oracle. Co-authored-by: Cursor <cursoragent@cursor.com>

Review 39142: replace nested-install assert with leaked-scope flush+log; add PerModuleResolveMemoScopeGuard (Drop flushes) around infer+pop. Document why module_name is excluded from BoundedMemoKey. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-07-17T01:59:46Z

Review 39147 — dashboard relay included only (not posted — add --post to comment on the PR); full stdout body was not available here.

Verified locally against likely scope concern: branch was behind still-owl-689 (#6780 fleet_intent memory ground), so the PR diff accidentally reverted unrelated dag/gunbc/fleet_intent.dag changes. Fixed: rebased onto current origin/session/still-owl-689; PR diff is now 5 files (memo + receipt tests only). Receipt (a) still green.

New HEAD: $(cd /home/briansrls/.worktrees/loyal-crab-450 && git rev-parse --short HEAD)

If review 39147 had additional findings beyond scope, please re-post with --post so they land on-thread.

— sent from loyal-crab-450

gunbai-bot · 2026-07-17T01:59:51Z

Correction: rebased HEAD is 9215b0a (prior comment had unexpanded shell fragment). — sent from loyal-crab-450

gunbai-bot · 2026-07-17T02:09:02Z

Review 39148 — APPROVE received; mild concern addressed (no code change):

Scope-extended env / bypass inertness: Within the install window, infer_scope.type_env is env.clone() (same Rc ptr as PerModuleResolveMemoScopeGuard::install(&env)), and peel_alias_once_for_field_access / resolve_node paths under infer_items+populate_output_provenance thread that same scope.type_env — not fresh TypeEnv allocations. Env-ptr bypass is for genuine sub-env divergence, not normal inference.

Whole-tree telemetry (operator release b on witness module): ~1787 bounded misses then hits climb linearly with ~100% hit rate (misses frozen); bypass did not dominate. Memo is active on the real hot path for that module — the remaining stall is caller-level superlinear resolve_node invocation count (documented on-thread), not memo inertness.

Will watch [resolve-memo-mod] bypass counters on clever-seal's main-434s re-price; escalation if bypass dominates is env-identity in key (already designed).

— sent from loyal-crab-450

Co-authored-by: Cursor <cursoragent@cursor.com>

…ecursive scrutinees The per-module resolve memo (#6779) had two grains sharing one fp cache: a bounded memo (resolve_node_bounded — its main within-module speedup claim) and a scrutinee memo on resolve_scrutinee_type_node keyed by a FULL-subtree surface fingerprint. On a self-recursive scrutinee (e.g. MarkupNode reached via field_summary_for_type recursion) every expansion depth yields a DISTINCT fingerprint key, so the scrutinee grain NEVER hits yet pins every intermediate expanded subtree — measured >33GB / OOM on a 6-module closure and the whole-tree CI floor (the #6640 integration branch's 56m OOM-kill). Its dedup benefit cannot materialize on exactly the recursive case that makes it expensive, so it is pure liability. Removed. The bounded grain is unaffected and keeps #6779's speedup. Verified: whole-tree compile-clean now completes bounded (~1.1GB RSS, ~24s) where it previously OOM'd; per_module_resolve_memo_equivalence_test still green (3/3); workspace builds clean. Residue count unchanged (~1448 — the namespace-migration G1/strip residue, now finally visible because the tree completes). Unblocks #6640 (namespace integration) and #6809 (census consolidation). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ecursive scrutinees The per-module resolve memo (#6779) had two grains sharing one fp cache: a bounded memo (resolve_node_bounded — its main within-module speedup claim) and a scrutinee memo on resolve_scrutinee_type_node keyed by a FULL-subtree surface fingerprint. On a self-recursive scrutinee (e.g. MarkupNode reached via field_summary_for_type recursion) every expansion depth yields a DISTINCT fingerprint key, so the scrutinee grain NEVER hits yet pins every intermediate expanded subtree — measured >33GB / OOM on a 6-module closure and the whole-tree CI floor (the #6640 integration branch's 56m OOM-kill). Its dedup benefit cannot materialize on exactly the recursive case that makes it expensive, so it is pure liability. Removed. The bounded grain is unaffected and keeps #6779's speedup. Verified: whole-tree compile-clean now completes bounded (~1.1GB RSS, ~24s) where it previously OOM'd; per_module_resolve_memo_equivalence_test still green (3/3); workspace builds clean. Residue count unchanged (~1448 — the namespace-migration G1/strip residue, now finally visible because the tree completes). Unblocks #6640 (namespace integration) and #6809 (census consolidation). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: Cursor <cursoragent@cursor.com>

…grounding memo (#6779) * feat(infer): per-module resolve memo (Option B) on still-owl-689 Within-module transient memo at resolve_node_bounded grain catches infer_items and populate_output_provenance re-grounding without the Option-2 persistent census roots that OOM'd. Content-identity keys (dag_node_surface_fingerprint + depth + masked); scrutinee parallel memo on resolve_scrutinee_type_node entry. Freed at module exit. Receipt (a): per_module_resolve_memo_equivalence_test. Co-authored-by: Cursor <cursoragent@cursor.com> * feat(infer): split resolve-memo counters into hit/miss/bypass Makes release (b) diagnostic: bypass counts env-ptr guardrail skips separately from fingerprint misses. Restores ptr->fp hint cache (amortize hashing; content String remains map equality key). Co-authored-by: Cursor <cursoragent@cursor.com> * fix(infer): pin Rc in ptr->fp hint cache (ABA-safe) Retain (Rc<Node>, fp) in node_surface_fp_cache so transient wrapper addrs cannot be reused with stale fingerprints during the module scope. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(infer): release RefCell borrow before memo compute (re-entrancy) Probe-then-compute-then-store pattern lets resolve_node_bounded recurse through the memo wrapper without panicking. Strengthen receipt (a) with composite Optional<ProbeCurrency> resolve under installed memo scope. Co-authored-by: Cursor <cursoragent@cursor.com> * test(infer): thread-local memo toggle for receipt tests Replace process-global GUNBC_PER_MODULE_RESOLVE_MEMO set_var/remove_var with PerModuleResolveMemoEnabledGuard so parallel cargo test cannot flake the byte-identity oracle. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(infer): panic-safe memo scope + module_name key invariant note Review 39142: replace nested-install assert with leaked-scope flush+log; add PerModuleResolveMemoScopeGuard (Drop flushes) around infer+pop. Document why module_name is excluded from BoundedMemoKey. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Brian Searls <briansearls1@gmail.com> Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

…ecursive scrutinees The per-module resolve memo (#6779) had two grains sharing one fp cache: a bounded memo (resolve_node_bounded — its main within-module speedup claim) and a scrutinee memo on resolve_scrutinee_type_node keyed by a FULL-subtree surface fingerprint. On a self-recursive scrutinee (e.g. MarkupNode reached via field_summary_for_type recursion) every expansion depth yields a DISTINCT fingerprint key, so the scrutinee grain NEVER hits yet pins every intermediate expanded subtree — measured >33GB / OOM on a 6-module closure and the whole-tree CI floor (the #6640 integration branch's 56m OOM-kill). Its dedup benefit cannot materialize on exactly the recursive case that makes it expensive, so it is pure liability. Removed. The bounded grain is unaffected and keeps #6779's speedup. Verified: whole-tree compile-clean now completes bounded (~1.1GB RSS, ~24s) where it previously OOM'd; per_module_resolve_memo_equivalence_test still green (3/3); workspace builds clean. Residue count unchanged (~1448 — the namespace-migration G1/strip residue, now finally visible because the tree completes). Unblocks #6640 (namespace integration) and #6809 (census consolidation). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: Cursor <cursoragent@cursor.com>

…grounding memo (#6779) * feat(infer): per-module resolve memo (Option B) on still-owl-689 Within-module transient memo at resolve_node_bounded grain catches infer_items and populate_output_provenance re-grounding without the Option-2 persistent census roots that OOM'd. Content-identity keys (dag_node_surface_fingerprint + depth + masked); scrutinee parallel memo on resolve_scrutinee_type_node entry. Freed at module exit. Receipt (a): per_module_resolve_memo_equivalence_test. Co-authored-by: Cursor <cursoragent@cursor.com> * feat(infer): split resolve-memo counters into hit/miss/bypass Makes release (b) diagnostic: bypass counts env-ptr guardrail skips separately from fingerprint misses. Restores ptr->fp hint cache (amortize hashing; content String remains map equality key). Co-authored-by: Cursor <cursoragent@cursor.com> * fix(infer): pin Rc in ptr->fp hint cache (ABA-safe) Retain (Rc<Node>, fp) in node_surface_fp_cache so transient wrapper addrs cannot be reused with stale fingerprints during the module scope. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(infer): release RefCell borrow before memo compute (re-entrancy) Probe-then-compute-then-store pattern lets resolve_node_bounded recurse through the memo wrapper without panicking. Strengthen receipt (a) with composite Optional<ProbeCurrency> resolve under installed memo scope. Co-authored-by: Cursor <cursoragent@cursor.com> * test(infer): thread-local memo toggle for receipt tests Replace process-global GUNBC_PER_MODULE_RESOLVE_MEMO set_var/remove_var with PerModuleResolveMemoEnabledGuard so parallel cargo test cannot flake the byte-identity oracle. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(infer): panic-safe memo scope + module_name key invariant note Review 39142: replace nested-install assert with leaked-scope flush+log; add PerModuleResolveMemoScopeGuard (Drop flushes) around infer+pop. Document why module_name is excluded from BoundedMemoKey. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Brian Searls <briansearls1@gmail.com> Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

…ecursive scrutinees The per-module resolve memo (#6779) had two grains sharing one fp cache: a bounded memo (resolve_node_bounded — its main within-module speedup claim) and a scrutinee memo on resolve_scrutinee_type_node keyed by a FULL-subtree surface fingerprint. On a self-recursive scrutinee (e.g. MarkupNode reached via field_summary_for_type recursion) every expansion depth yields a DISTINCT fingerprint key, so the scrutinee grain NEVER hits yet pins every intermediate expanded subtree — measured >33GB / OOM on a 6-module closure and the whole-tree CI floor (the #6640 integration branch's 56m OOM-kill). Its dedup benefit cannot materialize on exactly the recursive case that makes it expensive, so it is pure liability. Removed. The bounded grain is unaffected and keeps #6779's speedup. Verified: whole-tree compile-clean now completes bounded (~1.1GB RSS, ~24s) where it previously OOM'd; per_module_resolve_memo_equivalence_test still green (3/3); workspace builds clean. Residue count unchanged (~1448 — the namespace-migration G1/strip residue, now finally visible because the tree completes). Unblocks #6640 (namespace integration) and #6809 (census consolidation). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: Cursor <cursoragent@cursor.com>

…grounding memo (#6779) * feat(infer): per-module resolve memo (Option B) on still-owl-689 Within-module transient memo at resolve_node_bounded grain catches infer_items and populate_output_provenance re-grounding without the Option-2 persistent census roots that OOM'd. Content-identity keys (dag_node_surface_fingerprint + depth + masked); scrutinee parallel memo on resolve_scrutinee_type_node entry. Freed at module exit. Receipt (a): per_module_resolve_memo_equivalence_test. Co-authored-by: Cursor <cursoragent@cursor.com> * feat(infer): split resolve-memo counters into hit/miss/bypass Makes release (b) diagnostic: bypass counts env-ptr guardrail skips separately from fingerprint misses. Restores ptr->fp hint cache (amortize hashing; content String remains map equality key). Co-authored-by: Cursor <cursoragent@cursor.com> * fix(infer): pin Rc in ptr->fp hint cache (ABA-safe) Retain (Rc<Node>, fp) in node_surface_fp_cache so transient wrapper addrs cannot be reused with stale fingerprints during the module scope. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(infer): release RefCell borrow before memo compute (re-entrancy) Probe-then-compute-then-store pattern lets resolve_node_bounded recurse through the memo wrapper without panicking. Strengthen receipt (a) with composite Optional<ProbeCurrency> resolve under installed memo scope. Co-authored-by: Cursor <cursoragent@cursor.com> * test(infer): thread-local memo toggle for receipt tests Replace process-global GUNBC_PER_MODULE_RESOLVE_MEMO set_var/remove_var with PerModuleResolveMemoEnabledGuard so parallel cargo test cannot flake the byte-identity oracle. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(infer): panic-safe memo scope + module_name key invariant note Review 39142: replace nested-install assert with leaked-scope flush+log; add PerModuleResolveMemoScopeGuard (Drop flushes) around infer+pop. Document why module_name is excluded from BoundedMemoKey. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Brian Searls <briansearls1@gmail.com> Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

…ecursive scrutinees The per-module resolve memo (#6779) had two grains sharing one fp cache: a bounded memo (resolve_node_bounded — its main within-module speedup claim) and a scrutinee memo on resolve_scrutinee_type_node keyed by a FULL-subtree surface fingerprint. On a self-recursive scrutinee (e.g. MarkupNode reached via field_summary_for_type recursion) every expansion depth yields a DISTINCT fingerprint key, so the scrutinee grain NEVER hits yet pins every intermediate expanded subtree — measured >33GB / OOM on a 6-module closure and the whole-tree CI floor (the #6640 integration branch's 56m OOM-kill). Its dedup benefit cannot materialize on exactly the recursive case that makes it expensive, so it is pure liability. Removed. The bounded grain is unaffected and keeps #6779's speedup. Verified: whole-tree compile-clean now completes bounded (~1.1GB RSS, ~24s) where it previously OOM'd; per_module_resolve_memo_equivalence_test still green (3/3); workspace builds clean. Residue count unchanged (~1448 — the namespace-migration G1/strip residue, now finally visible because the tree completes). Unblocks #6640 (namespace integration) and #6809 (census consolidation). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot Bot marked this pull request as ready for review July 16, 2026 23:20

gunbai-bot Bot marked this pull request as draft July 16, 2026 23:23

gunbai-bot Bot changed the base branch from main to session/still-owl-689 July 16, 2026 23:23

gunbai-bot Bot force-pushed the session/census-pre-resolve branch 3 times, most recently from 51a5c84 to 84471e3 Compare July 17, 2026 00:57

gunbai-bot Bot changed the title ~~census pre-resolve (Option 2): ground global-bare census once at corpus level — fixes the O(K^2) resolve_node_bounded re-grounding hang~~ feat(infer): per-module resolve memo (Option B) — within-module type-grounding memo Jul 17, 2026

gunbai-bot Bot marked this pull request as ready for review July 17, 2026 01:10

gunbai-bot Bot force-pushed the session/census-pre-resolve branch from e7e22b9 to aafefa0 Compare July 17, 2026 01:15

gunbai-bot Bot force-pushed the session/census-pre-resolve branch from de0fac6 to 3431023 Compare July 17, 2026 01:31

gunbai-bot Bot marked this pull request as draft July 17, 2026 01:40

gunbai-bot Bot marked this pull request as ready for review July 17, 2026 01:43

Brian Searls and others added 6 commits July 17, 2026 01:59

fix(infer): pin Rc in ptr->fp hint cache (ABA-safe)

53525bd

Retain (Rc<Node>, fp) in node_surface_fp_cache so transient wrapper addrs cannot be reused with stale fingerprints during the module scope. Co-authored-by: Cursor <cursoragent@cursor.com>

test(infer): thread-local memo toggle for receipt tests

63d43af

Replace process-global GUNBC_PER_MODULE_RESOLVE_MEMO set_var/remove_var with PerModuleResolveMemoEnabledGuard so parallel cargo test cannot flake the byte-identity oracle. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot Bot force-pushed the session/census-pre-resolve branch from 3431023 to 9215b0a Compare July 17, 2026 01:59

gunbai-bot Bot marked this pull request as draft July 17, 2026 02:17

gunbai-bot Bot marked this pull request as ready for review July 17, 2026 02:23

briansrls merged commit 93a61d4 into session/still-owl-689 Jul 17, 2026

briansrls deleted the session/census-pre-resolve branch July 17, 2026 10:16

briansrls added a commit that referenced this pull request Jul 17, 2026

fix: cargo fmt after #6779 per-module resolve memo merge

ade111f

Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot Bot mentioned this pull request Jul 17, 2026

fix(infer): remove #6779 scrutinee resolve-memo grain — whole-tree compile-clean OOM #6813

Closed

briansrls added a commit that referenced this pull request Jul 17, 2026

fix: cargo fmt after #6779 per-module resolve memo merge

8227d56

Co-authored-by: Cursor <cursoragent@cursor.com>

briansrls added a commit that referenced this pull request Jul 17, 2026

fix: cargo fmt after #6779 per-module resolve memo merge

6b74f48

Co-authored-by: Cursor <cursoragent@cursor.com>

briansrls added a commit that referenced this pull request Jul 17, 2026

fix: cargo fmt after #6779 per-module resolve memo merge

c765812

Co-authored-by: Cursor <cursoragent@cursor.com>

Navigation Menu

Repository navigation

feat(infer): per-module resolve memo (Option B) — within-module type-grounding memo - #6779

feat(infer): per-module resolve memo (Option B) — within-module type-grounding memo#6779
briansrls merged 6 commits into
session/still-owl-689from
session/census-pre-resolve

briansrls commented Jul 16, 2026 •

edited by gunbai-bot Bot

Loading

Uh oh!

cursor Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

briansrls commented Jul 16, 2026 • edited by gunbai-bot Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Measured main-tree benefit (clever-seal whole-tree RNB attribution, 1090 modules)

Test plan

Head

Uh oh!

cursor Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

gunbai-bot Bot commented Jul 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

briansrls commented Jul 16, 2026 •

edited by gunbai-bot Bot

Loading