Repository navigation
fix(standards): io counts boundary operations — one owner, one statement, one boundary (#2841 contract, roadmap Phase 3) - #2844
Merged
Conversation
…ent, one boundary (#2841 contract, roadmap Phase 3) One io hit is an operation that moves data between the program and a system outside its own runtime, in a form an ordinary identifier cannot match (docs/io_rule_contract.md). 17 language rules edited on the five corollaries: - C1 ambiguity anchor: javascript bare path/https/database/sql out (probeIo's own `path` parameter was a hit), c calls paren-anchored, assembly in/out instruction-anchored, groovy lowercase generics out, dart types case-sensitive - C2 one owner: source/.read/.import/ATTACH are import's; close-family release forms are cleanup's in abap/c/cobol/cpp/fortran/livecode/lua/ matlab/perl/tcl/yacc (every one already claimed the token — the dual collapses, no mass lost) - C3 boundary: sqlite DML is in-engine computation; io is readfile/ writefile and the outbound dot-commands; abap's client DML stays - C4 one statement one hit + C5 job-local temps: jcl counts DD statements allocating an external target, DSN=&& never counts Sheet row io -> stated (5/68); signal_contract_audit baseline shrinks. Cross-language pins: tests/extraction/languages/test_io_contract_2841.py. Golden masters blessed; bless_scope: io + Amplified Rce (+ its weighted view) only, no aperture crossing. rosetta_audit 46/46 against the companion corpus branch: every language reads io = 3, closing the four open-defect cells (javascript 4, jcl 5, sqlite 9, tcl 7) and the two ledgered probe-idiom surpluses (c 4, lua 4). Closes #2841 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FCF3viid5wUh2dLqzfZ16v
squid-protocol
added
the
rosetta:rebless-owed
Golden masters re-blessed on the merged code (main's #2842 Mode E function-population fix + this branch's io contract). bless_scope vs main's masters: identical 482-diff scope — I/O and Network Boundaries + Amplified Rce (+ weighted view) only, newly parsed/excluded none. Extraction suite 6904 passed on the merge. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FCF3viid5wUh2dLqzfZ16v
Contributor
Owner
Author
|
Conflicts with main resolved by merge commit 6055522: main's #2842 re-blessed the same golden masters. Resolution was a re-bless on the merged code, never a hand-merge — |
squid-protocol
added a commit
to squid-protocol/keyword-rosetta
that referenced
this pull request
Sep 7, 2026
…(roadmap Phase 3) (#90) Every language now reads io = 3 (the median) under the stated contract (gitgalaxy docs/io_rule_contract.md). Cells and plants: - javascript main.js 4->3: bare `path` left the rule (C1); the collision probeIo's parameter documented is impossible now - tcl 7->3: source is import's, close is cleanup's (C2) - sqlite 9->3: DML is in-engine computation (C3), .read is import's; probe_io re-planted with writefile + .output beside readfile - jcl 5->3: one DD statement is one hit (C4), DSN=&& temps are job-local (C5); third external allocation planted (//DD5 DD SYSOUT=L) - c 4->3, lua 4->3: probe_cleanup's close idiom no longer double-counts Ledger: `identifier-name-token-collision` and `sqlite-dot-read-dual-import-io` resolved; `probe-idiom-collides-with- measured-signal` narrowed to haskell's safety_bypasses slice; batch4 loses tcl and the io token; new `io-contract-2841` entry. Orphan baseline regenerated. Report regenerated at full precision against fix/2841-io-contract: open-defect share 39 -> 35 of 2303 (1.5%), average consistency 92% -> 93%, io carries zero ledgered cells. rosetta_audit 46/46. Cross-repo: engine PR squid-protocol/gitgalaxy#2844 merges first; this branch goes green against engine main after. Claude-Session: https://claude.ai/code/session_01FCF3viid5wUh2dLqzfZ16v Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
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.
Phase 3 of the contract roadmap (
docs/contract_roadmap.md, epic #2812): theiofamily. Closes #2841.The contract
Full sentence, five corollaries, the 46-language audit table and the ledger dispositions:
docs/io_rule_contract.md. Sheet rowio→stated(5/68);signal_contract_audit.py --cigreen with a smaller baseline; cross-language pins intests/extraction/languages/test_io_contract_2841.py(CASES / COUNTS / single-owner checks / ReDoS detonations).What was wrong (the 4 open-defect cells, 4 different diseases)
\bpath\bcounted probeIo's ownpathparameter (ledgeridentifier-name-token-collision)//DD1 DD DSN=…,DISP=SHRfired twice (DSN + DISP=), and the&&tempteardown DDs counted.readdual with import (ledgersqlite-dot-read-dual-import-io)sourceclaimed by import+io,closeby cleanup+io (ledgerbatch4-dual-keyword-overlaps)17 rules edited. The close-family release forms leave io in abap, c, cobol, cpp, fortran, livecode, lua, matlab, perl, tcl, yacc — in every one, the language's cleanup rule already claimed the token, so each dual collapses to its owner with no mass lost (#2765's ENV/popdef shape). c and lua's ledgered probe-idiom io surpluses ("cannot be authored away") are authored away exactly there:
close(0)/io.close(conn)are cleanup's hits alone.Proof
rule_probe.py io allbefore → after: every corpus language reads io = 3 (was 4/5/9/7 on the four defects, 4/4 on c/lua). Crucible narrowings all attributed in the doc (assembly 230→90 is comment prose; sqlite 321→8 is DML; jcl 1590→884 is one-hit-per-DD).rosetta_audit.pyagainst the companion corpus branch: 46/46.bless_scope.py: 482 diffs, all in per-file sections, leaf keys onlyI/O and Network Boundaries+Amplified Rce(the proximity pair that reads io positions) + its weighted view; newly parsed/excluded: none.#2841 contractnote). ruff/mypy/dead-key audits: no new findings;audit_check.pyall clear.Filed on the way
remove(left io but c's cleanup doesn't claim it; the 11 cleanup rules that absorbed the close-family are presence-checked only).Cross-repo
Companion corpus PR: squid-protocol/keyword-rosetta#90 (
rebless/gitgalaxy-2841-io) — manifests for the 6 moved languages, the jcl//DD5 DD SYSOUT=Land sqlitewritefile/.outputre-plants, ledger retirements (identifier-name-token-collision,sqlite-dot-read-dual-import-io), narrowing (probe-idiom-collides-with-measured-signal→ haskell-only; batch4 loses tcl and the io token), and the newio-contract-2841entry. This PR merges first (rosetta CI checks out engine main); the corpus PR goes green and merges after, and push-to-main regen republishes the bias report.🤖 Generated with Claude Code
https://claude.ai/code/session_01FCF3viid5wUh2dLqzfZ16v