Answer: when it doesn't contain any useful code.
Update 2015-10-02 08.58: this post was updated to reflect the addition of code the metaPORE repository.
A discussion on twitter today revealed something which I find very disappointing:
A new paper by Greninger et al. (Rapid metagenomic identification of viral pathogens in clinical samples by real-time nanopore sequencing analysis) has been published in the journal Genome Medicine. The Methods contains the following line:
We developed a custom bioinformatics pipeline for real-time pathogen identification and visualization from nanopore sequencing data (MetaPORE) (Fig. 1b), available under license from UCSF at .
Reference #23 takes you to the metaPORE GitHub repository. At the time I initially wrote this post — and as the screen grab below shows — it contained zero code. Thankfully this has been changed and a set of Python and shell scripts are now available.
Maybe this was just some sort of error in scheduling the release of the paper and the code. However, journals and authors should understand that if a paper (or a pre-print) appears online and points to a code repository (or any other website), the expectation is that people should be able to visit the site in question and download code.