Repository navigation
Conversation
|
Independent source review of exact c776905: the faculty split passes source equivalence (231 functions, 52 declarations, 95 test annotations). Full scope is now clean: no prior epoch or engine-pin changes. File excess reduction 154999 -> 143321 is supported. [P2] Do not credit derived-bounds 159 -> 158 as removed debt. The parser at derived_bounds.rs:115-120 does not recognize pub(in ...) const. Existing PREFILL_TARGET_SECONDS (prompt/view.rs:18) and PER_MESSAGE_TEMPLATE_TOKENS (prompt/fit.rs:13) consequently disappear from scanning after this move. Meanwhile the rule-specific production_lines ignores file-level cfg(test), counting relocated AGENTIC_SURFACE_CEILING at tests/prompt_shaping/tool_surface_and_fit.rs:410. Net minus one is two lost production findings plus one misclassified test finding. Repair visibility parsing and test projection with the hygiene owner, then remeasure. Current #5227 migration fixes projection but still needs restricted visibility coverage. Logging limitation: four tracing calls and 40 moved probes inherit new child-module targets. Contents and probe classes are preserved; no exact old-target consumer was found. Preserve target parity through the explicit-target probe arm, or explicitly disclose this metadata change rather than claiming all observables unchanged. Read-only review; no local builds or deployment. Full PR remains on hold for honest bounds accounting and CI. |
…cargo_target is its own owner The flaky "kept" (canary 9c2e7b8, CI run 39444285785; #5232 c776905) is a real defect, not a test artifact. A reclaim pass released cargo's flock only by closing its descriptor. A flock belongs to the open file description, so a process forked while the pass held it (the core spawns git, engines and builds; tests fork in parallel) kept it until its exec, and the next pass read an idle tree as a live build. - HeldCargoLocks owns the held locks and unlocks each one on drop; a failed unlock is a probe, never silent. - Proving a tree is cargo's, holding its locks and taking it move to citizen_workspace_pool/cargo_target.rs (parent 993 -> 876 lines); callers unchanged (is_cargo_target re-exported pub(crate)). - Regression test: a dup standing in for the forked child stays open past the release, and the lock must still be free. Deterministic. Measured: a standalone rustc probe on the M5 (macOS, Darwin 25.5) showed a rival try_lock fails after close-only with a dup alive (false) and succeeds after an explicit unlock (true). continuum-core compile/test is left to CI (serving node, no build window); source-hygiene 43/43 at baseline. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit dd42301)
…cargo_target is its own owner (#5238) The flaky "kept" (canary 9c2e7b8, CI run 39444285785; #5232 c776905) is a real defect, not a test artifact. A reclaim pass released cargo's flock only by closing its descriptor. A flock belongs to the open file description, so a process forked while the pass held it (the core spawns git, engines and builds; tests fork in parallel) kept it until its exec, and the next pass read an idle tree as a live build. - HeldCargoLocks owns the held locks and unlocks each one on drop; a failed unlock is a probe, never silent. - Proving a tree is cargo's, holding its locks and taking it move to citizen_workspace_pool/cargo_target.rs (parent 993 -> 876 lines); callers unchanged (is_cargo_target re-exported pub(crate)). - Regression test: a dup standing in for the forked child stays open past the release, and the lock must still be free. Deterministic. Measured: a standalone rustc probe on the M5 (macOS, Darwin 25.5) showed a rival try_lock fails after close-only with a dup alive (false) and succeeds after an explicit unlock (true). continuum-core compile/test is left to CI (serving node, no build window); source-hygiene 43/43 at baseline. Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit dd42301) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
… by theme (pure move) Each nested test module becomes its own file; the loose prompt-shaping and verdict tests group by theme as on the owner split. include_str! paths gain two ../ for the added directory depth. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 28e811b)
…ure move) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 3b34e76)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit cfcf416)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit e3bda14)
…out of llm_deliberation_faculty.rs (pure move) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 5e3eced)
…culty.rs (pure move) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 9817768)
The depth rewrite skipped 'in super::' meant for pub(in super::..) but also matched 'for fact in super::perception_facts'. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 24e88f6)
…nly) ShedPreview, ContextComposition's framing fields and the GroundingPiece notice helpers sit in prompt/ now; private -> pub(in super::super) like their siblings. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit c776905)
Rebuilt as canary plus this PR's own commits (cherry-picked with -x). The old branch carried #5227/#5228's pre-squash history. Each changed count is traced to its cause in the comment beside it. cargo test -p source-hygiene passes 43/43 on a fresh per-worktree build. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo
d45e9d1 to
0f0d2cc
Compare
What this is
core/continuum-core/src/cognition/llm_deliberation_faculty.rswas 12,678 lines on canary. This splits it by owner intollm_deliberation_faculty/, with the same owner layout as the earlier split. Pure moves: no logic changes, no renames. Mechanical edits only: child modulesuse super::*;, private items widened topub(super)/pub(in super::super)where a sibling or the parent uses them,super::->super::super::in moved code one level deeper, parentmod child; use self::child::*;, test files start#![cfg(test)], andinclude_str!paths gain../per added directory level.It replaces #5226, which was stacked on the unmerged epoch branch (that branch rewrote ~1,600 lines of this file, so it could not be cherry-picked). This one is regenerated from canary's version of the file.
Canary-only items (not on the earlier split) were placed with their owner:
ContextComposition,GroundingPiece,GroundingPlan,Fittedand their impls go toprompt/fit.rs;ShedPreviewandshed_previewtoprompt/types.rs;latency_fill_cap,history_fill_cap,messages_within,PREFILL_TARGET_SECONDStoprompt/view.rs;FRONT_DROP_QUANTUM_DIVISORtoprompt/fit.rs. One deliberate difference from the earlier layout: canary's parent would still be 1,260 lines, over the 1,000-line rule, so the system-prompt composition (assembled_context_cost,render_assembled_context_within,compose_system*) moves to a newprompt/system.rs.Commits, one per owner: tests, tool_surface, request, verdict, prompt, reply_claim + hands_surface.
What moved where (lines)
llm_deliberation_faculty.rs(parent)hands_surface.rsreply_claim.rsrequest.rstool_surface.rsverdict/mod.rsverdict/contribute.rsprompt/mod.rsprompt/types.rsprompt/view.rsprompt/messages.rsprompt/fit.rsprompt/system.rsprompt/segments.rstests/mod.rstests/turn_bound.rstests/verdicts/{mod 165, act_verdicts 596, capacity_and_wire 852, window_feedback 659}.rstests/prompt_shaping/mod.rstests/prompt_shaping/{an_action_result_cannot_mute_a_citizen 172, grounding_never_mutes 99, the_budget_prices_the_surface_it_actually_sends 198, stable_prefix_first 139, turn_window 829, reply_reserve 871, prefix_and_grounding 419, tool_surface_and_fit 853, binding_and_history 455}.rsEvery file is at most 1,000 lines.
Verification
Not run locally (the machine that produced this is a serving node, so no cargo). What was checked locally: every new file parses under
rustfmt --edition 2021 --check, and a line-multiset comparison of the old file against all new files shows the only differences are the mechanical edits above (visibility,super::depth,mod/uselines,#![cfg(test)],include_str!depth).Verification of record is CI:
cargo test -p continuum-core --liband thesource-hygiene ratchets.🤖 Generated with Claude Code