Re: Debhelper for R packages
Hi Gordon,
On Thu, Sep 15, 2016 at 04:12:35PM +0200, Gordon Ball wrote:
> Moving forward with this, I propose to file an ITP for dh-r as a
> standalone package shortly, barring any objections. Since this shouldn't
> overlap with anything else, I would to go straight to unstable rather
> than going via experimental.
Perfectly fine. Can't wait to see this in unstable. :-)
> ## Repository
>
> Mirrored to alioth [1], and this is the canonical VCS url given in
> d/control.
>
> [1]: https://anonscm.debian.org/git/debian-science/packages/dh-r.git
Nice. I'd commit anything here in case I might find some enhancement.
> ## Vignettes
>
> A vignette dh sequence item is provided which should build vignettes,
> but it has to be explicitly requested. To use this you need to add a
> complete texlive environment to the build-depends, since this isn't a
> dependency of r-base-dev (won't someone please think of the buildds).
> This should also work with vignettes using knitr (but not rmarkdown,
> since it isn't packaged).
>
> %:
> dh $@ --buildsystem R --with vignette
Sounds very sensible.
> ## dh-make-R
>
> This has gained options to generate team-maintenance for debian-science
> or debian-med, and several styles of autopkgtests (and interactive help).
>
> dh-make-R --team science --tests run_unit_test,vignette
Not tested but from the pure reding this sounds very handsome. Thanks
a lot for this.
> ## convert-to-dh-r
>
> (Renamed from fix-deps.pm). This is only installed to /usr/share/dh-r,
> but will try and convert existing d/control files (although I wouldn't
> recommend using it in bulk without manual review, since it's pretty
> primitive).
OK, noted. I'll possibly give it a try even if manual conversion does
not seem to be that hard.
> ## Buildflags
>
> I was unable to inject hardening LDFLAGS options via the MAKEFLAGS
> environment option, so for the moment you'll still see
> hardening-no-bindnow. I don't think hardening is that high a priority
> for R, but this can be revisited later if someone has a better idea
> about how to do it.
I agree that hardening should not be a show stopper for this effort
and may be once it is available we get some more input.
> While I have done a certain amount of testing, doubtless there are still
> some pitfalls left. Caveat emptor.
I'm locking forward to get rid of cdbs in my packages. :-)
Again thanks a lot
Andreas.
--
http://fam-tille.de
Reply to: