Skip to content

Release seqproc 0.1.0 - #22

Merged
rob-p merged 65 commits into
mainfrom
dev
Aug 22, 2026
Merged

Release seqproc 0.1.0#22
rob-p merged 65 commits into
mainfrom
dev

Conversation

@rob-p

@rob-p rob-p commented Aug 22, 2026

Copy link
Copy Markdown
Contributor

Summary

Promotes the reviewed first-release implementation from dev to main.

The release candidate includes EFGDL 2, typed errors, named resources, FASTQ shard lists, stdin/stdout and interleaved I/O, three-lane inputs, graph planning/optimization, summary schema 1.13, release provenance, AVX2-tuned official x86_64 builds, and the final stdout broken-pipe policy.

This final boundary also:

  • resolves published antisequence = 0.1.0 from crates.io (checksum locked);
  • uses macos-15 for Apple ARM CI and cargo-dist builds;
  • updates the dependency audit to the registry release and schema 1.13.

Final gates

  • cargo fmt --all --check: pass
  • warning-denied all-target Clippy: pass
  • all-target tests and benchmark harnesses: pass
  • docs: pass
  • registry-resolved clean-room cargo package: pass
  • cargo publish --dry-run --locked: pass
  • cargo dist plan: pass

The macos-15 hosted job is currently queued after the same job remained queued on macos-14; Linux baseline, MSRV, and release-SIMD checks are green. We are using a bounded wait for hosted-mac capacity rather than blocking publication indefinitely.

rob-p and others added 29 commits August 21, 2026 19:53
Advance the git dependency to aa8cb0ad so seqproc builds include the
input-loss, bounded-match, and nested-graph error fixes from the
first-release review pass. All suites except the known diff_tests
legacy-output-arity item pass against this revision.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BqdQSxnbCGcF6BwGmswYTd
Independent review of the proposed release heads: verdict, verified gates,
release-blocking and post-release findings with file:line references,
compatibility-contract assessment, handoff corrections, and the status of
the 2026-08-21 fix pass with remaining open items.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BqdQSxnbCGcF6BwGmswYTd
Independent verification of the maintainer resolution pass at boundary
seqproc 4e0de31 / ANTISEQUENCE b5fecee: all claimed gates reproduced, the
SIMD library/executable split and matcher/finalization/optimizer fixes are
confirmed, and an A/B against the pre-fix build shows a large net matcher
performance win. Nine remaining blockers are listed (output-arity leak into
`run`, --unassigned2 truncation, finish-on-error materialization, the
disassembly-verified CPUID guard failure path, native-vs-v3 guard mismatch,
the per-read oracle-cache allocation, quadratic tie-break on long reads,
mixed-length Hamming seeding, and a dist-artifact provenance smoke), plus
cheap pre-tag improvements and the two open user decisions.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BqdQSxnbCGcF6BwGmswYTd
Advance the dependency to cd4a9e4, which replaces the dense per-read
reference-position cache with a sparse seed-hit-keyed map and bounds
equal-best tie resolution in the edit-search kernels. The 400k-barcode
whitelist workload drops from 51 s to 7.7 s single-threaded with
byte-identical output, and repetitive-read tie resolution is worst-case
independent of read length. Full suite green against this revision.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BqdQSxnbCGcF6BwGmswYTd
Records the mechanics of the per-read dense oracle cache and the
quadratic tie-resolution behavior, the sparse-map and windowed
early-termination fixes, and the benchmark and oracle-suite evidence.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BqdQSxnbCGcF6BwGmswYTd
Demultiplexed runs now reject explicit primary FASTQ outputs instead of
silently ignoring them, and unassigned outputs require exactly one target
per input lane in either direction. The SIMD byte-equivalence script
supplies both output lanes for the untransformed paired fixtures so the
strict run contract holds through the release gate, and it verifies both
lanes. build.rs applies -C target-feature=+/- overrides to the recorded
feature set so provenance and the runtime guard describe the exact
compiler contract, the release config documents that linker = "gcc" is
load-bearing for the ELF ISA note under rust-lld, the changelog records
the breaking topology changes and the graph-lifecycle propagation, and
the install docs are qualified on the coordinated crates.io publication.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BqdQSxnbCGcF6BwGmswYTd
Advance the dependency to 773e1ce: sticky finalization failures with the
typed GraphFinalizationFailed error, failure-path flushing of writers
that already streamed data (without creating or truncating outputs the
run never wrote), and a read-only fast path for per-read nested-graph
execution claims (reviewer implementation in 3617472 plus the maintainer
refinement keeping the hot-path poll Relaxed). Full suite, CLI workflow
tests, and the SIMD byte-equivalence gate pass against this revision.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BqdQSxnbCGcF6BwGmswYTd
Documents the cooperative closure of the second-pass blockers, the
verification matrix executed on the final boundary (ANTISEQUENCE 773e1ce,
seqproc pin 4566953), and the READY FOR RELEASE verdict: only the stdout
EPIPE policy decision, the 0.1.0 version confirmation, and the mechanical
release procedure remain. Also carries the maintainer's website
output-topology documentation and the load-bearing parser-boundary
comment.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BqdQSxnbCGcF6BwGmswYTd
Reviewer verification on report head ebf8aa3: the correction is
provenance-only, the schema-compatibility reasoning holds against the
1.13.0 enum, and formatting, the full suite, and clippy reproduce green
locally. The READY FOR RELEASE verdict stands; no further review pass is
required before the merge to main.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BqdQSxnbCGcF6BwGmswYTd
@rob-p
rob-p merged commit ec8639b into main Aug 22, 2026
9 of 10 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant