Repository navigation
Conversation
The Zenodo archive never changes, so the deflated bytes of each member sit at a fixed offset. Pin them, fetch one range per member, and let gzip check the CRC-32 and size the archive stores; the sha256 pin stays. Bead: pgp-bs9u.39
og dev fixture comphet writes the compound-het fixture byte for byte (CPython's Mersenne Twister for the seeded contig); og dev fixture trim-pharmcat trims a PharmCAT report; og dev evidence clashes and hg002-gvcf replace the public-evidence workflow's inline Python, streaming, with the covered positions kept as merged spans. Parity cases against the Python, and a golden for the site past a contig's end. Bead: pgp-bs9u.39
The personal-data scan, the image scan, the coupled-version check, the run-versions check, the URL expander of check-links.sh and the PGS Catalog answer readers, ported from the Python inside five scripts/ci scripts. Bead: pgp-bs9u.37
The 210-line checker the workflow wrote into a file is now an og dev verb; its pins table is data in scripts/ci/renovate-pins.tsv, where a missing file or a glob that matches nothing is an error. On the last saved log of main it prints the same lines as the Python, clean and with planted pins. Bead: pgp-bs9u.39
… their Python public-evidence.yml runs og dev evidence clashes and hg002-gvcf; the comphet e2e case builds its fixture with og dev fixture comphet. The comphet generator and the PharmCAT trimmer are deleted. Bead: pgp-bs9u.39
Bead: pgp-bs9u.39
Every field the trimmer keeps or drops, a non-object diplotype, a null allele, the recommendationDiplotypes fallback, non-ASCII text, and counts of 2, 0 and -1. Bead: pgp-bs9u.39
…next) Comphet seeded with 32, the trimmer drops activityScore, clashes needs three records, and renovate-check no longer flags an @ in a package name. Guard rust and parity must go red on this commit. Bead: pgp-bs9u.39
check-personal-data.sh, check-images.sh, gen-containers-config.sh --check-versions, check-links.sh and check-pgs-labels.sh call the og dev verbs instead of their Python heredocs, and each workflow job that runs them has og. New self-test controls: a scan of no text file, a stale image exemption, a versions.env line that is not NAME="value", and a JSON true read as an effect of 1. Bead: pgp-bs9u.37
Bead: pgp-bs9u.37
Bead: pgp-bs9u.37
Each self-test control and cargo test of the fixes must go red on this commit; the next commit reverts it. Bead: pgp-bs9u.37
# Conflicts: # CHANGELOG.md
|
Important Review skippedToo many files! This PR contains 110 files, which is 10 over the limit of 100. To get a review, reduce the PR to 100 files or fewer by splitting it into smaller PRs or changing its base branch. Upgrade to a paid plan to raise the limit. This review couldn't start because sufficient usage credits or metered capacity aren't available. Add credits or update usage-based reviews in the billing tab, then retry.
 |
…lanted fault The first red control stopped at clippy, so cargo test never ran and the PGS self-test step never started. This one plants the same five readings in a form clippy accepts; the next commit takes them back out. Bead: pgp-bs9u.37
Bead: pgp-bs9u.37
The dev-evidence-hg002-gvcf parity cases read ref.fasta and its .fai, which the *.fasta ignore rule kept out of the tree, so CI parity stopped with 'input not in tests/parity/inputs'. Bead: pgp-bs9u.39
# Conflicts: # CHANGELOG.md
# Conflicts: # CHANGELOG.md
The first red control stopped at clippy, so cargo test never ran. This one plants the same five pre-fix readings in a form clippy accepts, on top of the main merge; the next commit takes them back out. Bead: pgp-bs9u.37
Bead: pgp-bs9u.37
The *.vcf ignore rule kept the golden out.vcf of dev-evidence-hg002-gvcf/past-the-end out of the tree, so CI compared og with the oracle and reported the deliberate fix as a difference. Bead: pgp-bs9u.39
Select comphet-phasing-1-reads when its Rust generator changes. Keep the push paths and scope list aligned. Check HG002 truth records and the last valid reference block directly against the committed golden. Bead: pgp-bs9u.39
Guard must reject 1/1 reference blocks in evidence::tests::hg002_gvcf_keeps_truth_and_adds_only_valid_reference_blocks. Revert this planted fault after recording its red CI job. Bead: pgp-bs9u.39
…-c2b # Conflicts: # CHANGELOG.md # bin/copy_number.py # docs/rust.md # scripts/ci/e2e-case-paths.tsv # tests/parity/oracle.ref
…-z2e # Conflicts: # .github/workflows/guard.yml # CHANGELOG.md # docs/rust.md # rust/README.md # rust/crates/og-dev/src/lib.rs # scripts/ci/e2e-case-paths.tsv # scripts/ci/e2e-case-times.tsv # tests/parity/README.md # tests/parity/cases.tsv # tests/parity/oracle.ref
Collaborator
Author
|
@coderabbitai review |
Action not completed
|
…p JSON is accepted
…-z2f # Conflicts: # CHANGELOG.md # docs/rust.md # rust/crates/og-dev/src/lib.rs # rust/og/src/main.rs # tests/parity/cases.tsv # tests/parity/oracle.ref
…-z2d # Conflicts: # .github/workflows/guard.yml # CHANGELOG.md # rust/crates/og-dev/src/lib.rs # rust/og/src/main.rs # tests/parity/cases.tsv # tests/parity/oracle.ref
This branch has not been deployed
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
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.
A personal-data check could pass after scanning no text file. The Rust port also missed some Python matches and printed detected values into public logs. The latest fix restores ordinary Cyrillic text without admitting Unicode case differences.
The five shell checks now call
og dev. Their Python implementations are removed. Shell options and exception lists stay in place.check-personal-data.shog dev scan personal-datacheck-images.shog dev scan images,og dev versions coupledgen-containers-config.sh --check-versionsog dev versions check-runcheck-links.shog dev versions expandcheck-pgs-labels.shog dev pgs trait,label-match,metric-matchDeliberate fixes:
fancy-regex. Assertions participate in match selection. Necessary-literal checks skip impossible built-in rules.i. Literals, escaped members, positive and negative classes, and range interiors are checked. Case-sensitive Cyrillic names and ordinary Russian, Ukrainian and Serbian text work. Dot and category escapes remain supported. Such a refusal names the pattern file and line, says the cause, and says what to change: turn theiflag off for that part and list the exact spellings. It never quotes the pattern.Pipeline output filenames and formats do not change. FIFO behavior and the versions, images, PGS and links implementations are unchanged by this fix round.
The widened scope includes caller registrations and workflow wiring in
rust/og/src/main.rs,rust/crates/og-dev/{Cargo.toml,src/*},rust/Cargo.lock, and.github/workflows/{personal-data,guard,freshness,nextflow}.yml. The matcher adds pinned dependencies and lockfile entries. The containers parity fixture wraps its selector inprocess { ... }, as Nextflow requires. Parity adapters extract the removed Python from pinned Git sources. No Python source file is added.How it was checked
All inputs are public or synthetic. Finding tests run the removed scanner and compare sorted metadata, including duplicate findings. Redaction checks inspect stdout and stderr. Span tests compare match order and offsets with real Python interpreters.
The round-seven red-test commit fails all four controls: ordinary Cyrillic, pattern flag scope, indexed target validation before SKIP, and required interpreter failures. Guard replays it with
cargo test --no-fail-fastand requires every named failure.The converse verifier checks every differing BMP pair. Property-guarded endpoints must stop in all four entry modes. Admitted endpoints are tested with patterns that name no fold-class member under
i: dot, categories and their negations, unrelated classes, ranges, alternations, scoped flags and quantifiers. Case-sensitive Cyrillic patterns are compared too. Third case partners are discovered from real singleton relations on every interpreter. Differing named-class shapes must be refused. A span difference on admitted text under an accepted pattern fails verification. This is a bounded check, not a proof for unbounded patterns.U+1C89/U+1C8A are property-guarded. The test verifies that they stop before matching. Ordinary Cyrillic sentences scan clean in explicit/tracked and working-copy/index modes. A planted built-in finding inside Russian text remains exit 1. Case-sensitive custom names work. Case-insensitive names and ranges that name guarded fold classes return redacted exit 2.
Earlier red controls remain linked and replayed by Guard:
Guard also removes the BMP table and plants an incorrect word class and a swallowed matcher error. Each mutation must fail its named tests. The fixed source is restored after every control.
Earlier model evidence: 14,534 patterns, 9,314 accepted, 5,220 refused, 259 texts and 2,412,326 accepted pattern/text pairs. CI keeps the complete two-node/two-character slice: 598 patterns, 500 accepted, 98 refused, 43 texts and 21,500 pairs. Two span-fuzz seeds compare 298,627 and 292,152 pairs on Python 3.14. Unicode verification also runs Python 3.9/3.11/3.12/3.13/3.14/3.15 locally. Ordinary span iteration follows Python 3.15, including empty-text
\B. All 614 astral case sources and targets in the pinned UCD files remain guarded. The property table contains 830,004 scalars. The BMP table contains 32 differing pairs and now guards patterns separately.Time and memory on the same Mac,
/usr/bin/time -l, using a synthetic 20,000-line file with one finding per line:Both report 20,000 findings. The Rust output contains no planted value. These earlier measurements support no general speed claim.
Round-seven head:
65ba929fb967b785fc892628b799b60fbbd62a39. Local Rust 1.93 passes 136 unit tests and 11 repository tests, with one large model ignored by default. Clippy and rustdoc pass with warnings denied. Rustfmt passes. The corrected child-output assertion passes for both verifiers with missing and failing explicit interpreter paths.Six interpreters verify 1,243 candidate BMP pairs and the 32 checked-in differing pairs. The admitted-text converse compares 266,544 span rows. It covers 25 admitted endpoints and 26 property-stopped endpoints. All 150 differing named-class shapes are refused, including 108 third-partner shapes. The round-seven pre-fix replay reconfirms all four expected failures on this code, then restores the fixed sources.
The local oracle passes all 30 items in 10 cases. Its 15 self-tests and the comparator's 44 controls pass. Linux shadow controls remain CI checks. Personal-data, image and PGS shell self-tests and the links test pass. The single
containers-software-versionscase passes all four assertions for 128 synthetic processes. No real pipeline run is claimed. The public scan is clean for 2,439 text files: 6.21 s and 18,661,376 bytes peak RSS on this run.All upstream fast checks are green on that head: Guard with ci-ok, Rust, Lint, both Nextflow legs, Docs, Container Test and Personal data.
Guard's Rust job passes 862 workspace tests, all seven pinned scanner-control rounds, the missing-table control and both matcher mutations. It installs and requires Python 3.9, 3.12 and 3.15. Local replays also pass rounds one through seven and the missing-table control. The local incorrect-word-class mutation fails its named tests.
A final local control deliberately bypasses the property guard for U+1C89/U+1C8A. The revised converse then fails with an accepted-pattern span mismatch. Restoring the guard leaves the production source unchanged. Both verifiers also reject a deliberately unavailable required Python version.
READY
65ba929fb967b785fc892628b799b60fbbd62a39. Fork E2E scheduling and landing remain with the orchestrator. The PR stays open.CodeRabbit has no completed review or review threads. The earlier request was rate limited. The round-seven request followed the push after the shared queue cleared. It also answered rate limited. No repeat request was sent.
Beads
pgp-bs9u.37