Tracking: open PRs frozen pending maintainer review (stacks and merge order) #324

Context

PRs are paused at the maintainer's request: the recent wave caused merge conflicts with in-progress batching work, because new PRs reworked code that older PRs were already fixing. I will not open new PRs, and will not push or rebase existing ones, until I get the go-ahead.

This issue is an index of my open PRs so they can be triaged, ordered and batched (or closed) at the maintainer's pace. Nothing here needs action right now. Happy to close, squash or split anything on request.

Already merged (maintainer batches)

#284 (chimeric multimap), #289 (Batch1), #290 (Batch2), #291 (mappedFilter), #307 (WithinBAM, PE nM)

Wave of 2026-10-08 to 2026-10-10 (opened after the batches)

Many of these are stacked (base is another PR branch, not main). Merge order must follow the stack.

Stack A: alignment core / STAR parity (base main at #295)

#295 stitch annotated junctions, drop covered PE sub-alignments
 └ #301 never stitch same-diagonal seeds, PE junction walk
    └ #302 seed-overlap order, insertions, sjdb overhang
       └ #304 window formation / per-window dedup for long splices
          └ #311 last yeast differences
             ├ #312 sjdb insertion at mapping time
             └ #315 Aligned.toTranscriptome.out.bam record for record
                ├ #321 multimappers in STAR's window order
                └ #316 outFilterType BySJout / SJ.out.tab filtering
                   └ #317 twopassMode Basic as STAR
                      └ #320 genomeSAsparseD in seed search

Stack B: SAM optional tags (base main at #306)

#306 optional tags in --outSAMattributes order
 ├ #309 MC:Z / ch:A tags, --outSAMattributes All
 ├ #310 unmapped records fields
 └ #314 explicit XS implies outSAMstrandField intronMotif

Stack C: gene model from index (base main at #308)

#308 load gene model from index tables when no --sjdbGTFfile
 └ #313 mapping-time --sjdbGTFfile for TranscriptomeSAM tables

Standalone on main

• fix(quant): filter TranscriptomeSAM pairs as a unit, like STAR (#292) #293 TranscriptomeSAM pairs filtered as a unit
• fix(sam): paired-end nM is the pair's mismatch total, as in STAR #294 PE nM is the pair total (conflicting; likely superseded by Port STAR's WithinBAM output (supersedes #282); PE nM is the pair's #307, can close)
• docs(roadmap): mark 17.2-17.6 and 17.B done #296 docs(roadmap) status
• fix(sam): STAR's header text, from @HD VN:1.4 to the @CO command line #297 SAM header text as STAR
• fix(seed): padding sorts above every base in the SA binary search #300 SA binary search padding order (small, align/seed.rs)
• fix(seed): treat N-flagged SAindex prefixes as non-tight, as STAR's iSA1noN #303 N-flagged SAindex prefixes (small, index/sa_index.rs)
• feat(params): accept --outSAMtype BAM Unsorted SortedByCoordinate (both BAM outputs), outStd routed per output as STAR #305 --outSAMtype BAM Unsorted SortedByCoordinate
• Support comma-separated read file lists in --readFilesIn (several files per mate), with per-file read groups #318 comma-separated --readFilesIn lists
• feat(junction): read gzip GTF files (detected by magic bytes) #319 gzip GTF (small)

Umbrella / overlapping

• Close the alignment gap with STAR 2.7.11b (yeast + human, SE/PE, two-pass) #322 alignment gap with STAR 2.7.11b: overlaps Stack A heavily; should be either the single squashed version of Stack A or closed in favour of it, maintainer's call.
• feat(solo): reproduce cellranger count 10.0.0 outputs (--soloOutLayout CellRanger, clean-room) #323 (draft) cellranger count outputs: large, overlaps the solo PRs below; not for review now.

Older open PRs (before the batches)

• Stacked: fix(align): a soft-clipped variant of a pair is not a second locus #254 → fix(align): pair-subset dedup requires same chromosome and strand (stacked on #254) #271 (pair-subset dedup), perf(io): parallel order-preserving BGZF compression for BAM output #276 → perf(io): encode SAM/BAM records on the align workers (stacked on #276) #278 (parallel BGZF / record encoding)
• I/O: feat(io): optional pure-Rust parallel gzip input decoding (rapidgzip-core) #225, feat(io): detect input compression by magic bytes, optional bz2/zstd/xz #272, feat(io): optional paraseq FASTQ backend #273
• Quant: feat(quant): unspliced targets in TranscriptomeSAM, splicing-status tag and GeneSplicing counts for bulk total RNA-seq #283 (parts already landed via ## Port STAR's mappedFilter (supersedes #169, #252); TranscriptomeSAM crash fix and --quantMode validation from #283 #291)
• Solo / CellRanger: solo: cbMinP posterior threshold, oneExact guard, adapter-anchored geometry (replaces #150) #165, feat(solo): --soloOutRawBarcodes Observed — a CellRanger-shaped raw matrix (stacked on #173) #174, feat(solo): CellRanger behaviour by default on 10x geometry (changes default output, stacked on #175) #176, solo: --soloOutLayout CellRanger, CellRanger's output shape (changes output paths on 10x, stacked on #176) #178, solo: metrics_summary.csv under the CellRanger layout (stacked on #178) #179, solo: --soloCellFilter OrdMag, CellRanger's cell call (first item of #181, stacked on #180) #182, solo: count intronic reads by default on 10x (closes 30 of the 32 points vs CellRanger, stacked on #182) #184, feat(solo): STARsolo per-read SAM tags, CB_samTagOut and paired-end barcode reads (Phase 14.7) #226, fix(solo): port STAR's actual CellRanger4 clip rules #241, test(solo): a fixture the CellRanger flags can fail (#172) #243, fix(solo): port STAR's cbMinP, QSmax and oneExact for 1MM_multi barcode correction #274, feat(solo): optional CellRanger v3 .h5 count matrices via hdf5-pure #275
• Genome / index / aligner: genomeGenerate: libsais suffix-array builder, selected by --limitGenomeGenerateRAM #109, feat: STARlong binary + long-read window-coverage filter #116, aligner core: annotated-junction stitching, alignEndsType, in-recursion length penalty #146, clip: CellRanger4 3' poly-A trim #148, genome: --genomeTransformOutput SAM reports original coordinates #160, genome: --genomeType SuperTranscriptome #161, feat(degnorm): DegNorm-style transcript degradation normalization #232
• Tooling / docs: feat(cli): parameter files, genome version check, and a shell-run readFilesCommand #239, docs: record dependency decisions in DEPENDENCIES.md #246, bench: add a divan harness and the first hot-path benchmarks #248, test: yeast-tier differential in a minute, and a SAM agreement tool #253, docs(roadmap): record where the port stands, measured #255

Dependabot: #259 (noodles 0.116, already in #289, can close), #285, #286.

Hot spots (files touched by many PRs)

src/io/sam.rs, src/align/stitch.rs, src/align/read_align.rs, src/params/mod.rs, src/lib.rs. Any PR touching these should go through a batch rather than be merged on its own.

Proposal

When the maintainer is ready, I can: close the superseded ones (#294, #259, possibly #322 or the Stack A pieces), rebase a single stack at a time onto main, and re-run the STAR parity checks before each batch. Until then, frozen.