Summary

rustar-aligner --chimSegmentMin > 0 combined with --twopassMode Basic (and a --outFileNamePrefix value that doesn't end in /) aborts the run before writing any output:

[INFO  rustar_aligner] Chimeric detection enabled (chimSegmentMin=12)
Error: I/O error: No such file or directory (os error 2) (SAMPLE./Chimeric.out.junction)

Caused by:
    No such file or directory (os error 2)

No outputs are produced — Aligned.out.bam, Log.final.out, SJ.out.tab, and the chim file are all missing. The run is a hard failure.

This is a separable bug from #26 (--outFileNamePrefix treated as a directory):

• Without --twopassMode Basic: rustar treats <prefix> as a directory and writes all outputs (including Chimeric.out.junction) inside it. The prefix-as-dir behaviour is the --outFileNamePrefix ending in . is treated as a directory (STAR uses it as a literal string prefix) #26 bug, but it doesn't crash.
• With --twopassMode Basic: rustar still tries to write to <prefix>/Chimeric.out.junction, but in two-pass mode the chim writer fires in pass 1 before the dir-creating step that normal-output writing implicitly triggers, so the parent directory doesn't exist yet and the file open fails.

Pipelines that wrap STAR commonly add --chimSegmentMin via extra_star_align_args together with the rest of the standard STAR-fusion / arriba flag set, which includes --twopassMode Basic. Any nf-core/rnaseq run that enables both will crash silently (Nextflow surfaces only the non-zero exit; the error message points at a path issue, not a chim configuration issue).

STAR reference behaviour

STAR concatenates --outFileNamePrefix as a literal string when emitting <prefix>Chimeric.out.junction — no directory split, no parent-dir requirement, works regardless of whether the prefix ends in ., /, or any other character. See source/Chimeric.cpp and the file-opening glue in source/InOutStreams.cpp. The output path is built once at startup from P.outFileNamePrefix + "Chimeric.out.junction" and lives next to the other STAR outputs.

Reproducer

#!/usr/bin/env bash
set -euo pipefail
mkdir -p /tmp/rustar-mre-chim && cd /tmp/rustar-mre-chim

BASE=https://raw.githubusercontent.com/nf-core/test-datasets/626c8fab639062eade4b10747e919341cbf9b41a
curl -fsLO $BASE/reference/genome.fasta
curl -fsL  $BASE/reference/genes_with_empty_tid.gtf.gz | gunzip -c > genes.gtf
curl -fsLO $BASE/testdata/GSE110004/SRR6357072_1.fastq.gz
curl -fsLO $BASE/testdata/GSE110004/SRR6357072_2.fastq.gz

RUSTAR=ghcr.io/scverse/rustar-aligner:dev
STAR=community.wave.seqera.io/library/htslib_samtools_star_gawk:ae438e9a604351a4

mkdir -p idx-rustar idx-star
docker run --rm -v $PWD:/w -w /w $RUSTAR rustar-aligner --runMode genomeGenerate \
    --genomeDir idx-rustar --genomeFastaFiles genome.fasta --sjdbGTFfile genes.gtf \
    --sjdbOverhang 100 --genomeSAindexNbases 7
docker run --rm -v $PWD:/w -w /w $STAR STAR --runMode genomeGenerate \
    --genomeDir idx-star --genomeFastaFiles genome.fasta --sjdbGTFfile genes.gtf \
    --sjdbOverhang 100 --genomeSAindexNbases 7

# Identical flag set, two aligners. Both prefixes lack a trailing slash.
COMMON=(--readFilesIn SRR6357072_1.fastq.gz SRR6357072_2.fastq.gz --readFilesCommand zcat
        --runThreadN 4 --sjdbGTFfile genes.gtf --twopassMode Basic --runRNGseed 0
        --outSAMtype BAM Unsorted --outSAMattributes NH HI AS NM MD
        --outSAMattrRGline ID:WT_REP2 SM:WT_REP2
        --quantMode TranscriptomeSAM --quantTranscriptomeSAMoutput BanSingleEnd
        --chimSegmentMin 12 --chimOutType Junctions)

echo "=== STAR (--outFileNamePrefix STAR.) ==="
docker run --rm -v $PWD:/w -w /w $STAR STAR \
    --genomeDir idx-star "${COMMON[@]}" --outFileNamePrefix STAR.
echo "--- STAR outputs ---"
ls STAR.*Chimeric.out.junction STAR.Aligned.out.bam STAR.Log.final.out 2>&1

echo
echo "=== rustar (--outFileNamePrefix RUS.) ==="
docker run --rm -v $PWD:/w -w /w $RUSTAR rustar-aligner \
    --genomeDir idx-rustar "${COMMON[@]}" --outFileNamePrefix RUS. 2>&1 | tail -8 || true
echo "--- rustar outputs (none expected; run failed) ---"
ls RUS.* 2>&1 || true
ls RUS./ 2>&1 || true

echo
echo "=== rustar with trailing-slash prefix (workaround) ==="
mkdir -p RUSdir
docker run --rm -v $PWD:/w -w /w $RUSTAR rustar-aligner \
    --genomeDir idx-rustar "${COMMON[@]}" --outFileNamePrefix RUSdir/ 2>&1 | tail -3
echo "--- rustar outputs ---"
ls RUSdir/

Observed (verified, fresh paired run on commit 5f8ad08)

STAR with --outFileNamePrefix STAR.:

STAR.Aligned.out.bam
STAR.Aligned.toTranscriptome.out.bam
STAR.Chimeric.out.junction              # <-- chim file at top level, prefixed
STAR.Log.final.out
STAR.Log.out
STAR.Log.progress.out
STAR.SJ.out.tab
STAR._STARgenome/
STAR._STARpass1/

rustar with --outFileNamePrefix RUS.:

[INFO  rustar_aligner] Reading paired-end from ...
[INFO  rustar_aligner] Chimeric detection enabled (chimSegmentMin=12)
Error: I/O error: No such file or directory (os error 2) (RUS./Chimeric.out.junction)

Caused by:
    No such file or directory (os error 2)

--- rustar outputs (none expected; run failed) ---
ls: cannot access 'RUS.*': No such file or directory

rustar with --outFileNamePrefix RUSdir/ (parent pre-created):

Aligned.out.bam
Chimeric.out.junction       # <-- correct 14-column format
Log.final.out
SJ.out.tab

So the data layer is fine — Chimeric.out.junction content matches STAR's format when the path is reachable. Only the path-building / parent-dir-creation step is broken, and only when chim output combines with two-pass mode.

Suggested fix

The chimeric-output writer appears to construct its path as <outFileNamePrefix>/Chimeric.out.junction, treating the prefix as a directory unconditionally — but in two-pass mode it fires before the directory has been created (in non-two-pass mode, an earlier output writer happens to create it first, masking the bug). Either:

1. Route chimeric output through the same path-builder used for the main outputs (Aligned.out.bam, Log.final.out, ...). Those handle the bare-prefix case today, since #26 (--outFileNamePrefix treated as dir) causes them to write inside the dir but it does at least create the dir along the way.
2. Have the chim writer call std::fs::create_dir_all(parent_of(target_path)) before opening the output file. One-line fix in the chim-emission path.

The cleanest fix is #1 plus the broader #26 fix (treat prefix as a string concatenation, matching STAR). Then --outFileNamePrefix SAMPLE. yields SAMPLE.Chimeric.out.junction next to SAMPLE.Aligned.out.bam etc., matching STAR byte-for-byte.

Test plan

# After fix, both invocations should succeed and produce a chim file:
rustar-aligner --twopassMode Basic --chimSegmentMin 12 ... --outFileNamePrefix SAMPLE.
ls SAMPLE.Chimeric.out.junction   # currently does not exist; should after fix

rustar-aligner --twopassMode Basic --chimSegmentMin 12 ... --outFileNamePrefix dir/
ls dir/Chimeric.out.junction      # already works today; should still work

Why this matters

--chimSegmentMin is the standard way users enable fusion / chimeric detection on top of STAR — every STAR-fusion, arriba, and STAR-derived chimeric protocol begins with that flag. Most STAR wrappers (nf-core/rnaseq's extra_star_align_args, the STAR-fusion Snakemake wrapper, raw fusion-detection scripts) combine it with --twopassMode Basic and a STAR-prefix convention --outFileNamePrefix SAMPLE.. Today, any such invocation crashes against rustar before any output is written, with an error that doesn't obviously point at chim configuration.

Severity

Medium. nf-core/rnaseq doesn't currently pass --chimSegmentMin in its default test profile, so this isn't blocking the PR — but every user who adds it via extra_star_align_args will hit a silent failure mode that's hard to attribute without reading rustar source.

_{Filed during nf-core/rnaseq integration testing (nf-core/rnaseq#1855). All sibling issues from this exercise: author:pinin4fjords or grep for nf-core/rnaseq#1855.}