Skip to content

Problems reported upstream

Problems found in other projects’ data or tools while building irit, and what was reported.

FastQC accepts a FASTQ file cut off part way through its last line

Section titled “FastQC accepts a FASTQ file cut off part way through its last line”

Found while building leanqc’s test corpus; reproduced on FastQC 0.12.1 and 0.13.0 (2026-11-10).

A plain-text FASTQ whose last entry ends part way through its quality line is processed as complete: FastQC exits 0, writes a normal report and counts the cut-off entry as a full sequence (0.12.1 prints a bare null; 0.13.0 prints nothing with -q). A truncated .fastq.gz is correctly rejected, the case fixed for FastQC#45. FastQC’s reader catches files that end between lines, not within one.

Reported as FastQC#221, with a fix and an integration test in FastQC#222: reject an entry whose quality string length differs from its sequence (colourspace files excepted). FastQC’s full test suite passes with the fix; the new test fails without it. FastQC-Rust reproduces FastQC’s behaviour here by design. leanqc refuses such files.

Not reported: FastQC’s quality-encoding guess

Section titled “Not reported: FastQC’s quality-encoding guess”

FastQC treats a file whose lowest quality character is @ or above as Phred+64, so very high-quality Phred+33 data is misread. This was already raised as FastQC#212; the maintainer called it a known limitation of guessing, and FastQC 0.13.0 added a --phred64 option instead. leanqc uses a narrower rule (Phred+64 only when the highest character is between J and i) and has --phred.

GTDB R232 mock database doesn’t match its published checksum

Section titled “GTDB R232 mock database doesn’t match its published checksum”

Found by dblock on 2026-11-10.

r232_mockup.tar.gz (the small test database for GTDB-Tk R232) does not match the MD5 listed in the release’s MD5SUM.txt:

Value
MD5 in MD5SUM.txt 2c710f35d7e6296615bcf9026ee1c951
MD5 of the file served 59512f60fa48010647977460bb855666
Size 26,324,952 bytes (as served)
MD5SUM.txt last modified 2026-06-04 03:54 GMT
Tarball last modified 2026-06-09 11:09 GMT (data.gtdb.ecogenomic.org), 11:08 GMT (data.ace.uq.edu.au)

Both mirrors serve the same file. The two other files in the same folder (HOWTO.txt, test_genome_G021783475.fna.gz) and the full R232 package entry match their checksums, so the likely cause is that the tarball was rebuilt after MD5SUM.txt was generated. Anyone verifying the download (as GTDB-Tk’s docs suggest) gets a failure.

Reproduce:

curl -sL https://data.gtdb.ecogenomic.org/releases/release232/232.0/auxillary_files/gtdbtk_package/mockup_db/r232_mockup.tar.gz | md5sum
curl -s https://data.gtdb.ecogenomic.org/releases/release232/232.0/MD5SUM.txt | grep r232_mockup

dblock keeps GTDB’s published checksum in its catalog and refuses the file until one of the two is updated. Status: reported 2026-11-10 as Ecogenomics/GTDBTk#726.

gPlace: identical to pplacer, and an OOM kill before the --max_memory check

Section titled “gPlace: identical to pplacer, and an OOM kill before the --max_memory check”

cmc-aau/gPlace#5 (2026-11-10). Shares the gPlace checks: placements identical to pplacer for 59 genomes on the R232 ar53 and bac120 backbone trees, and the --max_memory floors (1.9G on class tree 8, 5.2G on the full bac120 tree). It also reports that on the full tree gPlace loads about 4.8 GiB before checking --max_memory, so a hard memory limit below that kills it instead of printing “needs –max_memory 5.1G”. Summary on GTDBTk#728.