Repository navigation
Fix replay of helper allocations above the single-width limit - #5991
Merged
Mister-Meeseeks merged 1 commit intoOct 28, 2026
Conversation
…imit Use the aggregate pool bound for certificate integer tokens while retaining input width limits. Cover producer-to-replay round trips at two and 512 maximum-width lifetimes, invalid aggregate values and false peak claims. Prepared with substantial OpenAI Codex assistance.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Change and mathematical scope
Two overlapping helpers of width
10^60each produce a valid capacity-2 * 10^60allocation, but the separate--verifyCLI rejects the emitted certificate withinteger token exceeds magnitude 10^60. PR #5763's large-width test covered the API, while its CLI replay fixture used small widths.Use a certificate-only integer reader bounded by
512 * 10^60, the largest pool the existing input schema permits. Check token length before integer conversion. Lifetime inputs retain their original bounds, and the occupancy, peak lower-bound and input-hash checks remain intact.This is a tooling repair with no new kappa, supplier admission, changed mathematical hypotheses, or selected-frontier change. Maintainer review and hosted CI remain outstanding.
Task and overlap
Started from canonical main
19fdb086ea96af9a9f87936a2b888e10af4ed810, priority revision 8, and refreshed before submission. The upstream head is unchanged. This addresses the independent tool-interface review handoff in Discussion #5752, surfaced through #1166. Theportable-replaycard (scope key831cc1a3c8f9ff01b197) is refined to this certificate serialization round-trip failure, rather than an environment mismatch.The concrete object is the helper allocator's emitted JSON certificate, within its existing domain of at most 512 lifetimes, width at most
10^60, and endpoints at most10^12. The predecessor is PR #5763, headcee2336bc3a20c586b1dbb55152d97a114c657a6, already incorporated into main. This adds independent CLI regression evidence and repairs its certificate reader. A refreshed scan of four live API pages covering 370 open PR bodies found no matching fix; Discussion and sparse local overlap coverage was partial.Validation
Validation on native Windows, CPython 3.12.10:
python -B -m unittest discover -s tests -p 'test_research_helper_allocation.py' -v: 9 tests pass.python -I -O -B: 9 tests pass.python -B scripts/research_selected_baseline.py:PASS_RESEARCH_BASELINE_BINDINGS_ONLY;git diff --checkpasses. Test temporary files used a workspace-local TEMP/TMP directory.No full scientific replay or full repository test suite was run; the selected-baseline check verifies bindings only.
Credit
Credit Ali Romman's PR #5763 and its retained PR #2655/#5717 predecessors. Prepared for josecarlosgonzalezroque-design with substantial OpenAI Codex assistance. Existing source licenses and predecessor attribution are retained.
Open next questions
The smallest useful follow-up is independent replay of the two-helper and 512-helper regression cases on hosted CI. No new research package or broader research claim is proposed.