fix(trimgalore): clean stale outputs on same-workdir retry

[pinin4fjords]

Summary

Three changes to the trimgalore module:

1. Stale-output cleanup at the top of the script block (the actual fix)

Add to both SE and PE branches:

rm -f *.fq.gz *.html *.zip *_trimming_report.txt

Makes the module idempotent on a same-workdir retry. Pattern is exhaustive over trim_galore's emit globs and cannot match input symlinks (those are *.fastq.gz, not *.fq.gz).

2. Pin cutadapt in environment.yml

bioconda::trim-galore=0.6.10 does not bound cutadapt, so the conda solver picks the latest. Add bioconda::cutadapt=5.2 to match the docker container build (quay.io/biocontainers/trim-galore:0.6.10--hdfd78af_2). Same pattern as the nf-core/cutadapt module's environment.yml.

3. Filter the python-version line out of the snapshotted log

The trim_galore log contains This is cutadapt 5.2 with Python X.Y.Z. The X.Y.Z is whatever python conda happens to resolve, which is not something the module is responsible for. Pinning python in environment.yml just to satisfy the snapshot is brittle (every patch release would need an env bump). Instead, filter that line out via findAll { !it.startsWith("This is cutadapt") } and drop it from the snapshot. Cutadapt version is still asserted via the separate Cutadapt version: 5.2 line in the log header. This matches the strategy used by the nf-core/cutadapt module's tests, where the comment is // python versions differ in the default conda env and container.

Why (1) is needed

When a job retries in the same workdir as a partially-completed previous attempt, an intermediate <prefix>_1_trimmed.fq.gz from the failed attempt can survive into the successful retry. The reads output glob *{3prime,5prime,trimmed,val}{,_1,_2}.fq.gz matches all three resulting files (_trimmed, _val_1, _val_2), and downstream consumers expecting 1-2 fastqs (notably fq/lint after #11227 added arity) fail.

Reported via nf-core/rnaseq users on 3.23.0+.

Related

• Paired-end Trim Galore outputs can cause fq lint after trimming to receive too many FASTQs rnaseq#1807
• fq/lint: constrain reads input to arity 1..2 #11227 (fq/lint arity)

Notes for reviewers

• Fix applied to both SE and PE branches symmetrically; SE is exposed to the same orphan risk in principle.
• No emit-glob change.
• Snapshot delta is just the removal of the This is cutadapt ... Python ... line in five places.
• Tested locally with docker; CI matrix is green on this branch.

🤖 Generated with Claude Code