Repository navigation
BS-27: the notation of a tooling module as language-1.2 source (#269) - #452
Conversation
…ing module The fifth slice of the front end as source (#269), on BS-24: the row (register 482 -> 483, the §31 table, the RP-07 check and its case), the scenario, the §17.23 prose, the CHANGELOG entry, the README's claim, the CI part, and the case, conformance_bs_27, which holds SurfaceTooling.printToolingModule, parseToolingModule, and declarationSpans to surface::tooling::{print_module, parse_module, declaration_spans} on hand-built vectors that reach every error and the limit on nesting at both depths, on the committed JSON files printed as modules and read and spanned back, on the committed tooling modules, and on seeded edits, with 20 planted defects each named by its vector. VERIFICATION.md records the first run: the evaluator agrees on 654 prints, 218 reads, and 229 spans, rustc on 5,098 runs, and every plant is detected. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bootstrap/src/SurfaceTooling.lex.tex over SurfaceGeneric: printToolingModule removes the spec and prints with printGeneric; parseToolingModule reads one object with parseGeneric and adds the implied spec in order of key; declarationSpans walks the members with a state machine over the tokens, reading a member with readFrom at depth 1 and an element of declarations at depth 2. Lean verifies the module and each root is eligible with exactly its declared effects, the printer's including recursion. The entrypoint, the lock, the three export roots, and the project README follow. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
BS-27 stacks on the combined landing of BS-22, BS-23, and BS-24 (#451), whose export, records and counts it takes; the merge is clean. The bootstrap lock follows the SurfaceTooling entrypoint on the merged tree. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The conflicts are generated bootstrap files only, taken from the line and regenerated in the commits that follow. The surface tooling part joins the Intel print lane that #450 requires of every BDD part. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With the surface tooling, the production export of the bootstrap project is 135,273,215 bytes, beyond the 134,217,728 it configured (LLS8002 in the certificate phase). The maintainer raised the limit to 536,870,912, which leaves room for the further components of the driver; SPEC.md §17.23, the CHANGELOG, and the BS-27 record say so. A semantic input: the lock and every package move in the regeneration that follows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The local disk cannot hold the production export and verify records. This workflow repeats the local regeneration on a clean checkout and uploads the patch; it is deleted in the commit that applies that patch. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Regenerated by regenerate.yml (run 43050104167) on this tree with the raised source limit; the temporary regen-452.yml is removed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… tooling packages Written by the count writer of #442 from LEXLEAN_COUNTS=fast runs of the nine parts on the regenerated tree, then validate-model --write. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…t record conformance_bs_27 passed whole with LEXLEAN_MEASURE_STACK=all on the regenerated tree and recorded the row of bootstrap/print_tooling_module (PD-08); VERIFICATION.md records the moved counts and how the tree was regenerated. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
afflom
left a comment
There was a problem hiding this comment.
The previous missing-product finding is superseded at exact034efd: print_tooling_module, parse_tooling_module and declaration_spans now have their four package files; SurfaceTooling build/verify records and the print stack row are committed. The bounded review closed the full980-blob delta from6b using tree inventories because the compare response capped at300 files. Producer43050104167 provenance and selected manifest/package metadata are consistent; no producer or test was repeated, and uploaded-artifact-to-commit byte equality was not independently established.
P2 — BS-15's total remains stale after the count rewrite. rust_package.rs constants imply327+182+2000=2509 header runs and308+137+308=753 render runs, totaling3262; the model/SPEC/CONFORMANCE/scenario still claim "all 3,190 runs". This is source arithmetic, not a new runtime measurement. Please reconcile the authoritative v0.4 statement and mirrors with the actual corpus; identical mirrors can still share a stale numeral.
The new VERIFICATION landing paragraph also names render_tooling_latex as the third added package, but that package already existed; the added package is declaration_spans. Please reconcile its assertion that every new crate fits the300-node evaluator bound with the unchanged bounded-subset constants (BS07=292, BS13=63, BS14=296, BS08=351) while totals increased by3. This review does not supply a new bound measurement.
The present-tense PR body still says the packages are absent and regeneration waits for451; update it for the current committed products and evidence limits. Existing owner retains these corrections and qualification. This remains a draft with limited draft-event CI, not full acceptance. Coordinate the inherited Cargo.lock advisory via #472; no duplicate resolver, regeneration or qualification run is requested. #269 remains open.
# Conflicts: # CHANGELOG.md
Part of #269: the authoring notation of a tooling module (
BS-27,surface/tooling.rs) as language-1.2 source, with rootsprintToolingModule,parseToolingModule, anddeclarationSpansofSurfaceTooling. With it,surface/tooling.rshas a certified twin.Draft, stacked on #451 (
BS-22toBS-24).SurfaceToolingis built from the generic notation ofBS-24: its printer, its reader, and its value readerreadFromstarted at a depth, which is the oracle'sgeneric::value.Design
printToolingModuleremoves thespecmember by structural recursion and prints the rest withprintGeneric, so it admitsrecursion.parseToolingModulereads one object withparseGeneric, refuses a value that is no object at its first scalar that is not white space and a writtenspecat the first occurrence of the word, as the oracle does, and adds the impliedspecin order of key.declarationSpanswalks the module's members with a state machine over the tokens, reading a member at depth 1 and an element ofdeclarationsat depth 2, so the limit on nesting falls where the oracle's does. Neither of the last two admitsrecursion. The decoder of a tooling module and the checker stay in the host.Evidence
Measured on the normative host on BS-24 over BS-23, BS-22, BS-21, and SL-13:
Lean verifies the module, and the eligibility report gives each root exactly its declared effects. The first runs found no fault of the module; VERIFICATION.md records what they did find (a preservation rule, the axiom policy of six definitions, and nesting vectors one level short of the limit).
Gates at the head
cargo fmt --check,validate-model(483 IDs),validate-spec-links(483 rows), clippy on the workspace with-D warnings, the 17 RP cases, and the locks current, on the merge with #451.CI is expected red until the regeneration
The three packages are not committed, so
conformance_bs_27stops at the stack-bound step after the differentials above have agreed. The regeneration owes them, the lock, the bootstrapexpected/records, the stack row ofprint_tooling_module, and the recount of the bootstrap count cases. It runs after #451 lands.🤖 Generated with Claude Code