chore: release v0.6.0

[nh13]

🤖 New release

• fgumi-bgzf: 0.5.0 -> 0.6.0
• fgumi-dna: 0.5.0 -> 0.6.0
• fgumi-tag: 0.5.0 -> 0.6.0
• fgumi-raw-bam: 0.5.0 -> 0.6.0
• fgumi-bam-io: 0.5.0 -> 0.6.0
• fgumi-metrics: 0.5.0 -> 0.6.0
• fgumi-sam: 0.5.0 -> 0.6.0
• fgumi-consensus: 0.5.0 -> 0.6.0
• fgumi-simd-fastq: 0.5.0 -> 0.6.0
• fgumi-sort: 0.5.0 -> 0.6.0
• fgumi-umi: 0.5.0 -> 0.6.0
• fgumi: 0.5.0 -> 0.6.0

Changelog

fgumi-bgzf

[0.6.0] - 2026-08-10

fgumi-dna

[0.6.0] - 2026-08-10

fgumi-tag

[0.6.0] - 2026-08-10

fgumi-raw-bam

[0.6.0] - 2026-08-10

Bug Fixes

• [breaking] Make consensus read downsampling deterministic (#727)

Documentation

• Name the tc tag in comments that still say pa (#695)

fgumi-bam-io

[0.6.0] - 2026-08-10

Bug Fixes

• Stream -o - to stdout instead of creating a file named - (#689)

fgumi-metrics

[0.6.0] - 2026-08-10

fgumi-sam

[0.6.0] - 2026-08-10

fgumi-consensus

[0.6.0] - 2026-08-10

Bug Fixes

• Count consensus reads, not templates, in --threads mode (#681)
• Restore single-strand consensus via --min-reads 1,1,0 (#683)
• [breaking] Make consensus read downsampling deterministic (#727)

Testing

• Pin record_consensus_pair to two emitted reads (#694)

fgumi-simd-fastq

[0.6.0] - 2026-08-10

Bug Fixes

• Bring the SIMD lexer under the crate-wide unsafe policy (#676)

Testing

• Compare long-record bytes, not just their length (#693)
• Sweep SIMD/scalar lexer agreement with proptest (#677)

fgumi-sort

[0.6.0] - 2026-08-10

Bug Fixes

• Report the thread counts each phase actually uses (#691)
• Stream -o - to stdout instead of creating a file named - (#689)

Features

• [breaking] Size the temp-file limit to the process fd budget (#705)

Performance

• Make compare bams faster than the sort it validates (#696)
• Break the k-way merge into its component costs (#706)

fgumi-umi

[0.6.0] - 2026-08-10

fgumi

[0.6.0] - 2026-08-10

Bug Fixes

• Count consensus reads, not templates, in --threads mode (#681)
• Bring the SIMD lexer under the crate-wide unsafe policy (#676)
• Report the thread counts each phase actually uses (#691)
• Scale the per-thread memory budget by the sort phase (#692)
• Compute supplementary TLEN instead of copying the mate primary's (#684)
• Restore single-strand consensus via --min-reads 1,1,0 (#683)
• Stream -o - to stdout instead of creating a file named - (#689)
• Compare a displaced sort-key run by digest instead of aborting (#700)
• [breaking] Make consensus read downsampling deterministic (#727)

Documentation

• Mark fgumi production ready as of v0.5.0 (#669)
• Keep the binary name out of per-argument help (#670)
• Name the accepted values for sort's boolean flags (#690)
• Stop denying that the pa tag was ever written (#682)
• Name the tc tag in comments that still say pa (#695)
• Clarify mate tags and update aligner recommendations (#717)

Features

• Accept on/off/1/0 for boolean flags via BoolishValueParser (#703)
• [breaking] Size the temp-file limit to the process fd budget (#705)

Miscellaneous Tasks

• Disable Miri isolation so the raw-bam proptest can run (#701)
• Retarget CodeRabbit config on measured review evidence (#675)
• Report scheduled-workflow failures as a tracking issue (#702)
• Gate the publish list per-PR and skip on exact version (#721)
• Gate PRs that touch the Miri-scoped crate (#712)

Performance

• Make compare bams faster than the sort it validates (#696)

Testing

• Sweep SIMD/scalar lexer agreement with proptest (#677)
• Pin reported consensus reads to records written (#685)
• Reap the cat children the streaming tests spawn (#716)
• Guard boolean-flag help across every command (#704)

Changed

• Breaking: consensus downsampling now selects which reads to retain by a hash of the read
  name rather than by shuffling a seeded random number generator. This affects simplex --max-reads, duplex --max-reads-per-strand, and codec --max-reads.

  The retained subset is now a pure function of the tag family, so it is reproducible across
  runs, thread counts, and execution modes. Mates share a read name and therefore a rank, so
  the two ends of a template rank together; codec --max-reads selects whole pairs and so
  always keeps or drops both ends, while simplex --max-reads (which caps the whole family by
  record count, fgumi#723) and duplex --max-reads-per-strand (which caps each end
  independently) can retain one end of a template without the other. Previously the surviving
  reads depended on how many families the
  consensus caller had already processed, so --threads N and the no---threads path could
  produce different consensus reads from the same input — measured at 566 of 77,804 records
  (0.73%) on one library at --max-reads-per-strand 2.

  Which reads are retained differs from previous releases for any run that sets a cap.
  There is no way to fix the reproducibility bug while preserving the old selection, since the
  old selection was the bug. Runs that do not set a cap are unaffected and remain
  byte-identical. Ports fgbio#1166.

Bug Fixes

• duplex --max-reads-per-strand 0 is now rejected at startup. It previously exited 0 having
  written an empty BAM, because a cap of zero empties every strand. simplex and codec
  already validated their equivalents.

Features

• Boolean command-line flags now additionally accept on, off, 1, and 0, alongside the
  existing true/false/yes/no/y/n/t/f. All spellings remain case-insensitive, and
  values outside that set are still rejected.
• Boolean command-line flags now render a true|false placeholder in help, so fgumi filter -h
  shows --memory-per-thread [<true|false>] rather than --memory-per-thread [<MEMORY_PER_THREAD>], which read as a request for a per-thread memory size.

---

This PR was generated with release-plz.

[coderabbitai]

Important

Review skipped

Ignore keyword(s) in the title.

⛔ Ignored keywords (2)

• chore(deps
• chore: release

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 555ed8a1-75cd-4d89-b3e5-45b0bcae840f

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

Note

Reviews paused

Use the following commands to manage reviews:

• @coderabbitai resume to resume automatic reviews.
• @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

• ▶️ Resume reviews
• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.12%. Comparing base (ade4853) to head (83bc264).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #671      +/-   ##
==========================================
- Coverage   94.12%   94.12%   -0.01%     
==========================================
  Files         181      181              
  Lines      109415   109415              
==========================================
- Hits       102987   102983       -4     
- Misses       6428     6432       +4     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[nh13]

@coderabbitai pause

[coderabbitai]

✅ Action performed

Reviews paused.