You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Test-only execution efficiency on release/v0.4, based on cf76090d473540f4008ec373d4c5ed96ad13cc01 (merged #450). In crates/conformance/src/cases/bootstrap/rust_package.rs, BS-15 constructs full-library Run arguments for every accepted, shaped and seeded render. Renders above the existing 16,384-byte evaluator limit are then excluded from within; compiled discards their full runs and constructs the existing 4,096-byte-prefix native runs again. The large full runs are never evaluated or compared.
Construct the same ordered evaluator/native vectors directly, materializing each small render once and only the existing bounded native input for large renders. Keep all manifest/render corpus generation, seeds, counts, accepted-manifest package::package/recipe assertions, negative fixtures, two lowered roots, 21 plants, native differential, certification limits and exported-package byte comparisons unchanged. No orchestration, timeout, host gate, normative language or compiler identity change.
Acceptance
Source review establishes exact old/new input and expected-outcome equality, including the 16,384-byte boundary and 4,096-byte prefix.
A focused regression compares old and new assembled vectors across both profiles, committed-style expected files and recipe inputs.
Required formatting, lint, tests and the complete release gate remain necessary on the final tree. Do not report a measured speedup until hosted timing evidence exists.
Coordination
#442 owns count-report/writer tooling. Preserve its count reporting; this issue changes only how render run vectors are materialized, not their membership or count assertions. #272 remains the broader complete certified-packager product work and must stay open. No other current PR claims this optimization.
#459 was externally merged as f8854bf, with ordered parents 6a02d2c + afe6c67. The optimization and regression remain unchanged: actual net scope is rust_package.rs plus verification history, with the same 797 native/241 evaluator renders and 556 avoided unused encodings (source deductions, not new timings).
The completed head run43493536296 passed all70 jobs and772 steps, including same-run honesty; exact repro43493536304 passed. However, its b5390de5 tree differs from actual90732a41 because the actual merge also includes #517/#462 history. That successful earlier-tree run cannot alone qualify the composition. Keep #457 open until the already-existing actual-tree CI43532545468 completes and is reviewed; actual repro43532545412 passed. No new dispatch, producer or repeat of a completed gate is requested. #272 remains open, and accepted actual history must reach the existing #405 forward lane. This is a postlanding acceptance gap, not a premerge approval.
Retained source/equivalence reviews and the final-tree regression satisfy the full issue scope: ordered vectors and expected outcomes,16384boundary/4096prefix, both profiles and supplied-file/recipe inputs, with all corpus/count/assertion/negative/native/certification/export guarantees preserved. No measured current speedup is claimed and no completed check was repeated. This supersedes the pending actual-tree gate in comment6869133510.
Closing this narrow delivered issue explicitly because #459 landed on release/v0.4. Broad #272 remains open; #405 still owes actual accepted forward history and its own final-tree acceptance. Receipt: issue457-final-acceptance-20261114/receipt.json.
Activity
test(bootstrap): avoid discarded BS-15 full-library runs
Merge accepted BS-21 release base into BS-15 efficiency candidate
afflom commentedon Nov 14, 2026
#459 was externally merged as f8854bf, with ordered parents 6a02d2c + afe6c67. The optimization and regression remain unchanged: actual net scope is rust_package.rs plus verification history, with the same 797 native/241 evaluator renders and 556 avoided unused encodings (source deductions, not new timings).
The completed head run43493536296 passed all70 jobs and772 steps, including same-run honesty; exact repro43493536304 passed. However, its b5390de5 tree differs from actual90732a41 because the actual merge also includes #517/#462 history. That successful earlier-tree run cannot alone qualify the composition. Keep #457 open until the already-existing actual-tree CI43532545468 completes and is reviewed; actual repro43532545412 passed. No new dispatch, producer or repeat of a completed gate is requested. #272 remains open, and accepted actual history must reach the existing #405 forward lane. This is a postlanding acceptance gap, not a premerge approval.
afflom commentedon Nov 14, 2026
Delivered through merged PR #459. Independent full-body acceptance and postlanding review5903110468 now cover actual f8854bf / tree90732a41fa749a83859577c2ea3fb52df343e6b6. Exact actual CI https://github.com/UOR-Foundation/LexLean/actions/runs/43532545468 attempt1 passed all70 jobs/772 steps, including same-run honesty; exact reproducibility https://github.com/UOR-Foundation/LexLean/actions/runs/43532545412 passed.
Retained source/equivalence reviews and the final-tree regression satisfy the full issue scope: ordered vectors and expected outcomes,16384boundary/4096prefix, both profiles and supplied-file/recipe inputs, with all corpus/count/assertion/negative/native/certification/export guarantees preserved. No measured current speedup is claimed and no completed check was repeated. This supersedes the pending actual-tree gate in comment6869133510.
Closing this narrow delivered issue explicitly because #459 landed on release/v0.4. Broad #272 remains open; #405 still owes actual accepted forward history and its own final-tree acceptance. Receipt: issue457-final-acceptance-20261114/receipt.json.