Re: [Fedora-r-devel-list] Building RPMs for Fedora with circular "suggest" dependencies
On 06/15/2012 08:27 AM, William Emmanuel S. Yu wrote:
Hello,
I am new to this list and in the process of building some RPMs for R packages. I have
been rolling my own packages for a few years now and post them for public use. I am a
heavy user of R2spec with spec file tweaks of course. Thanks for the author for making
life easier.
Every so often, I encounter the circular dependency issue with "suggest". For
example, I am building R-reshape that depends on a bunch of packages including R-iterator
and R-foreach. Unfortunately, R-iterator and R-foreach both suggest each other.
I used to be able to force ignoring suggests but this causes problems with check and test
and skipping check and test isn't exactly what I want to do either.
I would like to know what the "proper" way of building RPMs for the both of
these R packages.
Hah.
From the perspective of a scientist or engineer, the 'proper' solution
would be to have a dependency _TREE_ instead of a dependency
plate-of-pasta.
In short, your problem looks insane because the underlying reality is
insane.
http://article.gmane.org/gmane.comp.lang.r.devel/21496
includes my statement of the problem, ca. 2009.
http://article.gmane.org/gmane.comp.lang.r.devel/21597
for a stab at a solution. It's not elegant, but I think it'd work.
I've frankly been out of the loop on where Pierre-Yves is going on this
issue for some time. I'm sure he's got some comments.
- Allen S. Rout