[rkward] rkward/rbackend/rpackages/rkward: rebuilt 'rkward' package

Thomas Friedrichsmeier thomas.friedrichsmeier at ruhr-uni-bochum.de
Thu May 24 06:44:39 UTC 2018


Hi,

On Wed, 23 May 2018 23:41:47 +0200
meik michalke <meik.michalke at uni-duesseldorf.de> wrote:
>  - ver.R is missing in the sources because it must be generated
> during build to always show the correct version string
> 
>  - but the file must still be listed in the collate field of the
> DESCRIPTION file in the sources because package build will fail
> otherwise
> 
>  - also .rk.app.version must be exported in the NAMESPACE file
> 
>  - so cmake is being called to generate ver.R locally, but the file is
>    actually only needed to trigger the entries in DESCRIPTION and
> NAMESPACE
> 
> is that about correct? if so, then a local dummy of ver.R with an
> arbitrary exported function .rk.app.version() should also do the
> trick, making the cmake call unnecessary. instead of just a check for
> the file the script could actually just generate one itself. it would
> be replaced with a proper one if RKWard was *really* built, so
> there's no harm done.

.rk.app.version is a character vector with the (full) version string,
so RKWard can detect whether the correct version of the library has
been loaded. So the cmake call is still needed. We used to have it in
git, but that keeps causing annoying merge conflicts, which is why I
removed it, eventually.

I know you like to use the ready-made .debs, but arguably it is a good
idea to attempt a regular build before any commit, anyway. (And that's
easy enough of debian systems, too).

BTW, I'm thinking about installing the R package to some
rkward-specific data-dir, instead of the system-wide R library
location. Then any worries about loading the wrong version (and thus
ver.R) will become obsolete, and it will also become a bit easier to use
R with several R installations on the same machine.

Not something for 0.7.1, though.

> also, i suppose it would be even better to use the (adjusted) content
> of longer roxyPackage script instead of just roxygenise(), because
> that also updates all the other files where info like version number
> or packaging date are stored (like in the *-package.R file, where the
> latest build still claimed it was rkward 0.6.3 from 2014-10-30).

I see. Would you like to adjust scripts/roxygenize.sh, accordingly?

Regards
Thomas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://mail.kde.org/pipermail/rkward-devel/attachments/20180524/9da73da1/attachment.sig>


More information about the rkward-devel mailing list