Repository navigation
Conversation
…er the corpus census DESIGN §6's bare-minimum-cost standing rule: a proven cost-shape defect — a copied accumulator, a quadratic fold — is always fixed, regardless of the realized n. This is both, in one expression, over a population that is not small and does not stay fixed. WHAT IT WAS. `deduplicate_identities` decided membership by rescanning the survivors it had already kept (`any(unique, prior => prior == identity)`) and extended them by copying them (`concat(unique, [identity])`). Both halves are O(n) per element, so the fold is quadratic in comparisons AND quadratic in list construction. WHY n IS NOT SMALL HERE. The argument is a whole-corpus rustc diagnostic census: the two persisted receipts carry `genesis_5006ddc6_raw_error_diagnostic_identities` and `receipt_1_eced9d24_raw_error_diagnostic_identities`, and the population is the emitted compiler's error identities — it grows as the self-host corpus grows. It is reached from `canonical_identity_set` (every phase row, every unplaced/codeless partition) and from `census_population_is_duplicate_free` (twice per receipt in `receipt_population_coherent`), so the whole series ratchet pays it repeatedly. WHAT IT IS NOW. One traversal carrying the two questions it was conflating in one list: `seen` (a `Set<String>`) answers "has this identity been kept", `unique` answers "in what order". The `std.graph` DFS accumulators are the same shape for the same reason. BEHAVIOUR IS UNCHANGED, deliberately and at both grains that are observable. Order: an identity is appended exactly at its first occurrence, as before — and every caller except `census_population_is_duplicate_free` sorts afterwards through `canonical_identity_set`, which only counts. Multiplicity: the function still collapses repeats rather than refusing them, so the duplicate WALL stays exactly where it was — upstream, in `census_population_is_duplicate_free` and `receipt_population_coherent`, which is the distinction `a_duplicated_census_population_is_not_silently_collapsed_to_a_set` exists to hold. EXECUTED EVIDENCE, not a typecheck. Eight enrolled witnesses over the real persisted series and the duplicate walls pass on this tree, run individually through `gunbc run --claim-run --source-root dag --source-root src/v2 --entry dag/test/claim/self_host_compile_phase_frontier_witness_test.dag --function <claim>`: `a_repeated_population_entry_is_refused_rather_than_collapsed`, `a_duplicated_census_population_is_not_silently_collapsed_to_a_set`, `a_receipt_whose_refused_population_repeats_a_path_is_refused_by_coherence`, `the_same_receipt_without_the_repeat_is_coherent`, `current_persisted_compile_phase_frontier_holds`, `the_census_digest_is_invariant_under_reordering_of_either_population`, `the_persisted_genesis_census_is_fully_attributed_to_its_phases`, `phase_boards_are_invariant_under_population_reordering`. The first four are the discriminating REDs for this change: collapsing a duplicated population, or refusing one that is clean, reddens them. THIS IS A COST-SHAPE REPAIR, NOT A FLOOR UNBLOCK. It is landed on its own terms because §6 requires it, not to buy headroom for any claim. The separate question — that ten required-floor claims each recompute the whole series ratchet, which is authored duplication under §2 and wants one provider at their least common ancestor — is a restructure and lands separately, so its design review does not hold this defect hostage. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W377Pq4Tp5eQjGBXtPQEoM
Set equality is not the acceptance test for a set-backed deduplicate. A dedup that returns the SET's ordering produces the same MEMBERS in a different ORDER: identical for a caller that counts, byte-drifting for a caller that renders. The previous commit asserted order was preserved; this one proves it, and proves the assertion can fail. THE FIXTURE IS CHOSEN SO THE TWO ANSWERS CANNOT COINCIDE. First-occurrence order over ["b", "a", "b", "c", "a"] is ["b", "a", "c"]; every sorted or set-ordered spelling answers ["a", "b", "c"]. A population that happened to arrive already sorted would let an ordered-equality assertion pass while discriminating nothing, which is why the subject is authored rather than read off the live series. THE MUTANT IS THE WHOLE CHECK, and it was run rather than described. Planting `|> sort_by(identity => identity)` on the accumulator's result — the set-ordered variant — turns this arm RED, and turns RED a probe comparing the new function against the PRE-CHANGE algorithm over both live populations (`genesis_5006ddc6_raw_error_diagnostic_identities`, `receipt_1_eced9d24_raw_error_diagnostic_identities`). Unmutated, both pass: the new function answers the old function's exact ordered sequence on the real corpus census, not merely its member set. Four arms, in one run: A pass, B pass, C fail, C2 fail. WHERE THE ORDER CAN AND CANNOT ESCAPE. `deduplicate_identities` has exactly two consumers in the corpus: `canonical_identity_set`, which sorts the result and therefore erases order, and `census_population_is_duplicate_free`, which counts it. The projection consumer — `gunbc.design_ledgers compile_phase_frontier_blocks`, which renders committed bytes into docs/design-ledgers.md — reaches it only through the sorting one, and that order-independence is separately executed by `the_census_digest_is_invariant_under_reordering_of_either_population` and `phase_boards_are_invariant_under_population_reordering`, both passing here. So the projected bytes cannot move under this change by construction, which is a stronger statement than one byte comparison over one population would have been. The arm is therefore not a belt on either current caller. It holds the contract for the NEXT consumer — the one that would not survive a silent reordering, and that would discover it as generated-artifact drift at a distance from the diff that caused it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W377Pq4Tp5eQjGBXtPQEoM
… two compiler_tests failures on this PR's run are main's, inherited through a merge ref pinned at 06:47Z against bb96afa The merge ref CI built for f3839b2 was pinned at push time against main as it then stood (bb96afa), which carried #9886's stale mirror. That is the exact subject #10017 repaired at 06:56Z, nine minutes after this PR's run started. Evidence, checked rather than assumed: - main at bb96afa, run 33596615712: 'test result: FAILED. 642 passed; 2 failed' with render_rust_applied_type_routes_qualified_base_through_leaf_name and shell_service_unmodeled_output_key_refuses -- byte-identical counts and identities to this PR's rust-unit-tests failure. - #10017's own body names 'regen FAIL generated surface drift: compiler_tests.rs, std_realization_schedule.rs' reproduced on 4059156 and bb96afa -- the same two files this PR's build job reported. - This branch changes two .dag files and no Rust: git diff origin/main...HEAD over compiler_tests.rs and std_realization_schedule.rs is empty. Merging rather than rebasing per the squash-merge policy, and pushing after main's last move so the merge ref is recomputed against the repaired base. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W377Pq4Tp5eQjGBXtPQEoM
…control it is missing main landed a competing repair for the same defect while this PR waited for a floor slot (`|> sort_by` then compare against the last kept element). Mine used a Set for membership and `list_push` for the accumulator. Both fix the quadratic scan; I argued mine additionally fixed a copied accumulator, and I measured that argument rather than asserting it. THE MEASUREMENT REFUTES MY PATCH. Three variants over an amplified probe — ~720 identities with every one duplicated, ten dedup passes per run, two rounds, same host: mine 112.1s / 110.3s, main's 110.6s / 127.3s, main's-with-`list_push` 110.9s / 137.6s. Round one spans 1.5% and round two is dominated by host noise in the wrong direction. `concat(unique, [identity])` is not a measurable O(n) copy in this interpreter, so §6's bare-minimum-cost argument does not distinguish the two implementations. Forking a landed fix on an unmeasured preference is the thing this project keeps catching, so main's implementation is taken whole and `dag/gunbc/self_host_compile_phase_frontier.dag` is byte-identical to `origin/main` in this commit. WHAT SURVIVES IS A GAP IN THAT FIX, NOT A COMPETING ONE. main's fold compares only against the LAST KEPT element, which is sound exactly because the input was sorted first — equal identities are adjacent. The sort is therefore a correctness precondition, not a presentation choice, and nothing in the corpus pins it: delete `|> sort_by(identity => identity)` and the fold silently stops collapsing duplicates separated by any other identity while still collapsing adjacent ones. MEASURED, NOT ASSERTED, INCLUDING THE PART THAT DECIDES WHETHER IT IS WORTH LANDING: A new witness against main's implementation PASS B same witness, `sort_by` deleted FAIL C the four pre-existing duplicate-wall witnesses vs that mutant ALL PASS C is the finding. `a_repeated_population_entry_is_refused_rather_than_collapsed`, `a_duplicated_census_population_is_not_silently_collapsed_to_a_set`, `a_receipt_whose_refused_population_repeats_a_path_is_refused_by_coherence` and `current_persisted_compile_phase_frontier_holds` all stay green with the sort removed, so this arm is the only thing in the corpus that reddens on it. THE SUBJECT IS CHOSEN SO THE MUTANT CANNOT PASS: `["b", "a", "b"]` holds a duplicate that is not adjacent in encounter order — 2 under the sorted fold, 3 without the sort — and the all-adjacent cases sit beside it as the positive control the mutant still passes, which is what makes the first arm discriminating rather than the set being merely two examples. IT ASSERTS DUPLICATE SURVIVAL RATHER THAN OUTPUT ORDER, deliberately. Encounter order is unobservable through both consumers (`canonical_identity_set` sorts its result, `census_population_is_duplicate_free` reads only the count), so an order assertion would pin a property no caller can distinguish — and the previous revision of this branch did exactly that, which main's change would have reddened. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W377Pq4Tp5eQjGBXtPQEoM
briansrls
marked this pull request as ready for review
September 3, 2026 01:25
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Contributor
|
Closing as superseded — this PR has no remaining content, and fixing its CI would be work on an artifact that should not land. Its one contribution is already in main. The 28-line sort-precondition control for GitHub's diff view is misleading here, which is why this is worth spelling out. The three-dot diff against the merge base shows "1 file changed, 28 insertions" — a clean, reviewable-looking addition. But the merge base predates And the file itself is stale in a way that matters. Main's copy of So there is nothing to rescue. One method note kept from that lane, because it generalizes: comparing per-claim floor — sent from crisp-ibex-710 |
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.
Auto-opened by session-dashboard for session
stern-lark-508.Pushing to
session/stern-lark-508advances this PR.Worker attestation
Before flipping this PR to ready for review, confirm each item:
npm test,cargo test) and the result.Closes #Ndirective.Summary
TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.
Test plan