Keeping pace with upstream

Goal

Produce byte-identical fastqc_data.txt and summary.txt for the same inputs as Java FastQC, so the two can be used interchangeably.

Equivalence testing

UPSTREAM.toml pins the tracked Java FastQC version. A nightly CI job checks for new releases and opens a GitHub issue when one appears. Every commit is tested against stored Java reference output across 18 test cases, covering text files (byte-identical), HTML structure, SVG charts (normalised), and pixel-level PNG comparison. See the equivalence report for details.

The only known differences are in chart rendering — Rust bundles Liberation Sans instead of system Arial, and uses resvg + tiny-skia instead of Java2D for PNG output, so antialiasing and text positions differ slightly — plus deliberate bug fixes documented in the changelog, such as using the specified Phred+33 encoding for BAM/SAM input instead of inferring it (#6).

Upstream contributions

Pull requests submitted to s-andrews/FastQC to port improvements upstream:

• #185 — Bundle Liberation Sans font (merged). Eliminates system font dependency and aligns chart text metrics between Java and Rust.
• #186 — Update integration test snapshots for the bundled font.
• #187 — Trim trailing empty rows from Adapter Content when --min_length is used.
• #210 — Use the Phred+33 encoding fixed by the BAM/SAM format instead of inferring it from the data (issue #209).

Features ported from upstream PRs (not yet merged):

• #161 — HTML report design refresh. Available in the Rust rewrite as the modern template (--template modern).

Coming soon (code awaiting clean-up and PR creation):

• pditommaso/FastQC — fork by Paolo Di Tommaso. Achieves 2-3x speed gains.