[Bio-packaging] testing out guix

[Ricardo Wurmus]

Ben Woodcroft <b.woodcroft at uq.edu.au> writes:

> And since I'm here, I notice bedtools is up to .24 on github, but .22 in 
> guix (which you wrote). There are some syntax changes though e.g. my 
> false bug report
> https://github.com/arq5x/bedtools2/issues/257
> What is the update procedure given there's a few tools that rely on this?

That’s bad.  I think that a change like that a) should not be enabled by
default, and b) deserves bumping the major version number as it’s
breaking other tools.

We could simply update the “bedtools” package and hope nothing breaks.
When other packages depending on the bedtools command line tools (rather
than includes or libaries) do in fact break because of this we could
offer different versions of bedtools (though it’s hard to justify for a
minor version).

We can also just refrain from updating if we’re afraid of breaking
things.  Instead, a user can inherit from “bedtools” and create their
own “bedtools-bleeding-edge” package:

    (define-public bedtools-bleeding-edge
      (package (inherit bedtools)
        (source ...) ; the new origin and hash
        ...))

Does .24 offer any substantial improvements that would justify updating
and potentially breaking things?  Or can we just wait for future
versions?

~~ Ricardo