Repository navigation
feat(checker): replay conditional bodies the kernel closes by case split - #337
Merged
Merged
Conversation
Statements without a theorem that the kernel closes by the two case-split rules (split on a goal condition, split on an inner condition) carry only the rule name in the record. The checker now replays them from the source when the function body is a multi-line conditional: the body is read by indentation into one parenthesised term, let-bindings are substituted, a field of the result is taken from the branches, and the goal is split on its first condition with the negation removed, then, if that fails, on the first atom of a connective. In the boolean branch of an algebra node the case pattern replaces the scrutinised argument. A place the checker still leaves on the kernel's word now names the kernel rule in its reason. New probe family if-body: two honest files replay with code 0, two forged records stay at code 3. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PNHrA3rG7FTWhB11E7pjDQ
New plan flang/proof/kernel-word.fscript with shortcuts kernel-word:corpus, kernel-word:stdlib and kernel-word:probes. For every file of the set it runs the kernel with a record, gives the record to the checker with per-statement output (a library file together with the records of the modules it imports), and reports how many statements the kernel proved, how many of them the checker replayed, why the rest stays on the kernel's word (a node with premises, or a kernel rule without moves), and for each of the seventeen kernel rules how many proved statements name it and how many of those were replayed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PNHrA3rG7FTWhB11E7pjDQ
After the verdict of the proof report, check --proof now gives the record to the independent checker (next to the binary, or named by FLANG_CHECKER) and prints one line: how many of the proved statements the checker replayed and how many stay on the kernel's word. Without a checker, or when the checker refuses the record, the line says that nothing was checked independently. The exit code does not change. The change lives in the hand-written runtime and reaches the binary with the next print. The trust ceiling of the whole path is rewritten: the checker grew by 133 code lines and the runtime by 112. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PNHrA3rG7FTWhB11E7pjDQ
ADR-0072 names the weak spot of the trust story: where the checker answers code 3 the verdict rests on the kernel alone. It sets the target that a statement counts as proved only when the checker replays it, with kernel-only verdicts reported apart; the path to it class by class with the measured counts of each class; and what "the kernel is proved" means here: the trusted base is the checker, the Lean model and the host, not the kernel. The checker page describes the new line of check --proof and the census plan. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PNHrA3rG7FTWhB11E7pjDQ
The four files of the if-body probe family get their rows in the proved share ledger, the English word "option" enters the file name words, and the counts in prose follow the tree: checker lines, flang files, C lines. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PNHrA3rG7FTWhB11E7pjDQ
ADR-0072 went to the parallelism decision on the trunk. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PNHrA3rG7FTWhB11E7pjDQ
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PNHrA3rG7FTWhB11E7pjDQ
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.
Verified on the tree of dev plus this branch (8722693): the checker test set passes, the provability verdict is green with all four checks, the proved-share ledger and the rule tables agree with the tree, pre-push is green, commit messages follow the convention.
🤖 Generated with Claude Code