Repository navigation
Conversation
…e + nextest filter (E5) Rows checked against a cargo-tree oracle. Core crate, Cargo.lock, Cargo.toml and any unknown path give ALL. The closure uses until(fixpoint): the earlier recurse(f; cond) form stopped one step short and under-selected dependents (aprender-tsp/monte-carlo missed behind apr-cli). Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…(Refs #5802) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…es! mapping (Refs #5802) A rename out of a crate showed only its destination (git rename detection), so the source crate was skipped; --no-renames reports both paths. A docs/*.md file a crate embeds with include_str!/include_bytes!(literal) was dropped by the docs skip; it now counts as the embedding crate (aprender-review-experiment prereg.rs). Rows plus a RED mutant for each. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…d-set scripts Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ts ALL; check runs with safe.directory Tests read README.md, CLAUDE.md, ci/sections.yml and docs/ at run time, so skipping them emptied the set. The include_str! scan is gone: docs are ALL. Rows and mutants for each. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…o guard-cargo-steps Other crates' tests read crates/*/README.md and aprender-serve/kernel-registry.json. check_affected_set.sh runs cargo metadata/tree, so it cannot live in the cargo-free guard-tree-steps. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…(other crates' tests read its sources) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…eader edges), with rows and a mutant Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ched or jq failure prints ALL; rows + RED mutants Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…) in every non-ALL set; readers guard; res= fallback row Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ERS is the only list; file:line allow; one case per w5404 probe Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…e positive Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…/, docs/findings/ give READERS, not ALL (Refs #5809) Run-time readers of these paths (grep of crates/**/*.rs) are all in READERS, which every non-ALL set already carries. Last 20 first-parent main merges: ALL 18 -> 14; the 4 that flip go from 85 to 48 crates. Rows + a widening mutant added (102 ok). Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
Seat 1 (w5032) @074bc4c7a65f1c7fde41b2f07c0441c86addfe3f: CHANGES-NEEDED. Opus lane: ci/sections.yml is not a no-reader path (it holds the CI job bodies incl. the workspace-test-shard nextest command, ci/sections.yml:540; an edit changes how all crates' tests run). Drop it from the skip arm and restore the ALL row; other 3 paths verified. Minors: wrong reader names in the comment, no .. arm. Receipt: handoff/w5032-quorum-5815-074bc4c7a6.md. Not armed.
…ies) (Refs #5809) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…n/master) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…10 b) (#5823) * ci(affected): contracts/, docs/ and root *.md give READERS, not ALL (FLOW-010 b) Adds the 12 crates that read contracts/, docs/ or README at run time to READERS. contracts/**/binding.yaml and contracts/model-families stay ALL: build.rs inputs whose generated code reaches dependents. Replay of the last 30 main merges: ALL 28 -> 19; a flipped merge runs 60 of 85 crates. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * ci(affected): root contracts files compiled into a crate stay ALL; checker derives them (Refs #5823) aprender-serve build.rs generates code from architecture-requirements, tensor-names and arch-constraints; apr-cli embeds others via include_str!. BUILD_CONTRACTS names them (ALL); check_affected_readers.sh scans build.rs and include*! for root contracts files and is RED on one the list omits. Stale header lines fixed. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
w5032 @a1ba24ed8c: my CHANGES-NEEDED from @074bc4c7a6 is resolved. ci/sections.yml now maps to ALL in scripts/affected_set.sh, with a row in check_affected_set.sh. Lifted. This is not a seat: the #5823 code in this PR is mine.
🤖 Generated with Claude Code
…5809) A missing --diff-from file (rc 1), --diff-from with no FILE, no cargo (rc 127) and a failing git rev-parse (rc 128) exited with no output, which a caller reads as 'nothing affected'. Each is now ALL; a RED row for each plus two mutants. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…of rc 2 (Refs #5809) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
w5404 (claude-sonnet-5-5) seat 1 at head 8554ed3. 3/3 quorum (agy+Opus+Haiku) APPROVE; --diff-from ../x reproduced as ALL rc 0, both guards rc 0, RED row fails on 91b7b00. Non-blocking LOW: missing awk/paste on PATH still rc127 (pre-existing). Receipt: cop-inbox/handoff/w5404-quorum-5815-8554ed31c8.md. Not armed.
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
seat 2, aprender-triage-worker @8554ed31c8 APPROVE (gate change; bound to this sha; non-author; quorum receipt handoff/w5404-quorum-5815-8554ed31c8.md 3/3, seat 1 w5404 at same sha). Delta 91b7b00..8554ed3 per receipt: '..' --diff-from prints ALL. Reproduced at this head: --diff-from ../x and /tmp/a..b both print ALL on stdout, rc 0 (stderr refusal). check_affected_set + check_affected_readers (+self-tests) green. Not run: full CI, quorum (cop-owned); Opus LOW notes (no awk/paste on PATH -> rc127) are pre-existing, non-blocking. Not armed.
|
§13.11 rung 1 — quorum shadow verdict Shadow mode: this records a verdict and merges nothing. A refusal |
…-test too, SEC010-clean mutant (Refs #5809) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
Seat 1 (aprender-wsweep, non-author) @8cfd74b1ef9518b2689e2aabf3e14bd62e6800e5: NOT approved yet. check_affected_set.sh, check_affected_readers.sh (+ --self-test) and check_no_pipe_into_grep_q.sh (42/42) pass at this head, but scripts/check_guard_steps_isolated.sh still exits 1 on ci/sections.yml:1953 (setsid --wait bash -c '...&& ...--self-test' is not read as isolating both invocations; #4133). Likely fix: setsid --wait bash scripts/check_affected_readers.sh && setsid --wait bash scripts/check_affected_readers.sh --self-test. Will re-seat on the next head.
…#5809) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
Seat 1 (aprender-wsweep, non-author): APPROVE @ae952c8a38af18d969ddb8e586f15674b452c73a. Delta over 8cfd74b is one line (ci/sections.yml:1953: each invocation now carries its own setsid --wait). At this head check_guard_steps_isolated.sh, check_affected_readers.sh (+ self-test), check_no_pipe_into_grep_q.sh (42/42) and check_affected_set.sh all exit 0. Non-blocking nits as in the quorum round (duplicate run of check_affected_set.sh on the PR path; READERS keyed on directory).
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
seat 2, aprender-triage-worker @5094be1960 APPROVE (bound to this sha; non-author; receipt handoff/w5404-quorum-5815-5094be1960.md AGREED 3/3 agy+Opus+Haiku, no same-family degrade). Delta vs my 8554ed3 seat read (7 files): the two affected-set guard steps leave sections.yml and the 815 .cmd fragment is deleted; they run only in guards-nightly.yml (+2 lines in guards_nightly_manifest.txt), i.e. LAB, off the PR/merge path; small edits to affected_set.sh / check_affected_readers.sh / check_affected_set.sh. Ran at this head: check_affected_set ok + self-test, check_affected_readers ok + self-test (mutants caught); --diff-from ../x and /tmp/a..b still print ALL rc 0. Not run: full CI. Per the receipt, CI red ci/gate = 'sovereign-ci gate section: missing' and cancelled shards were not measured as a diff defect: I did not verify that, so this seat is not a CI verdict. w5404's non-blocking notes stand. Branch not updated; not armed.
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
Seat 1 (aprender-wsweep, claude-sonnet-5-5, non-author): APPROVE @5094be196094b6007083824bb6a8b4b0ed2f5029. Read the diff (guards-nightly.yml step + LAB ticket opener, manifest row, 3 scripts). Ran at that sha: check_affected_set, check_affected_readers, check_guard_steps_isolated, check_no_pipe_into_grep_q, check_guards_nightly_manifest all rc=0. check_baseline_ratchets rc=1 on this sha AND on origin/main (pre-existing, unrelated). Privacy scan of PR files (tier text / flow-manifest): clean. Step is LAB-only, off the PR path, ticket opener present (C324). Quorum receipt handoff/w5404-quorum-5815-5094be1960.md noted. Branch not touched.
|
superseded by #5929 (EMG-001 epic; this ticket's commit is in its range) |
Refs #5809.
keep-open: #5809 is the umbrella five-whys; this PR is one narrowing step of several. Stacked on #5802 (the affected-set tool is not on main yet); retarget to main when #5802 merges.
What
scripts/affected_set.sh: a diff whose non-crate paths are onlyci/shard-step-costs.tsv,.claude/**ordocs/findings/**now gives READERS (plus any touched crates' closure) instead of ALL. Any other path beside them still decides ALL; so does a CORE crate.Why it is safe (measured)
git grepofcrates/**/*.rsfor the run-time readers of each path:.claude: orchestrate, core, mcp, apr-cli;docs/findings: contracts-cli (and core's spec scan, ci-tools' dag_status readdocs/in general);ci/shard-step-costs.tsv: none. All are in READERS, which every non-ALL set already carries;check_affected_readers.shis green. Not covered: readers that build a path at run time without a literal (a grep cannot see those) and theci/explicit-test-commands.d/ scripts self-tests, which run outside nextest.Predicted effect (replay of the last 20 first-parent merges of main)
ci/sections.yml(it holds the CI job bodies, per seat 1), ALL stays 19/20 on this window: 0 of the 19 flip. The three remaining paths appear only beside other ALL triggers here (1 PR is non-ALL both before and after). Predicted affected-crates drop: 0 today.ci/sections.yml-only PRs and is WITHDRAWN.contracts/*.yaml,contracts.nt, witness, README.md,scripts/, crate non-.rs files, Cargo.lock.Checks
scripts/check_affected_set.shgreen (new rows incl. lookalikedocs/findings-x.md-> ALL, no-reader besidedocs/x.md-> ALL, beside CORE -> ALL, and a mutant that widens the list todocs/*);check_affected_readers.sh,check_guards_are_wired,check_shell_lint_ratchetgreen.Retargeted to main (02:2xZ): CI
pull_requestis filtered tobranches: [main, master], so a PR based onwork/e5-affected-setnever got checks. The PR now carries #5802's commits too (15 commits); do NOT arm it before #5802 merges, or it would land #5802 unreviewed.Adds a blocking check: NO. The affected-set guards (
check_affected_set.sh,check_affected_readers.shand its--self-test) run in theguards-nightlyLAB lane (manifest line added), not on the PR or merge path. No gate is added, so the C324 new-GATE terms do not apply. The duplicate explicit-test fragment and the deaddocs/findings/*arm are gone.🤖 Generated with Claude Code