Gentoo Archives: gentoo-science

From: James Cloos <cloos+gentoo-science@×××××××.com>
To: gentoo-science@l.g.o
Subject: Re: [gentoo-science] g-cran master and portage master incompatible
Date: Wed, 24 Aug 2011 01:25:33
In Reply to: Re: [gentoo-science] g-cran master and portage master incompatible by Donnie Berkholz
1 >>>>> "DB" == Donnie Berkholz <dberkholz@g.o> writes:
3 JC> I find that g-cran and portage master branches are incompatible.
5 DB> Ever do anything with this? Just tried g-cran myself and I'm hitting
6 DB> it too, on portage 2.2.
8 No, I just spent some time thinking about the problem.
10 I hat to sound to critical, given that it was written as a gsoc project,
11 but the lack of real Manifest files made g-cran highly unreliable, even
12 when it worked at all. A common problem I had was failed downloads, due
13 to network issues. Unlike with other packages, after such a file one
14 has to manually remove the g-cran downloads before portage can download
15 a working copy. And you can't use ebuild(1) to debug build issues with
16 c-cran ebuilds; only emerge(1) ever worked.
18 There were also some dependency bugs. Both unnecessary and
19 unfulfillable deps, and also some missing deps. I doubt that the
20 latter can be reliably fixed, but the former can be filtered out.
22 Given that osuosl is a cran mirror, I think a better solution would be
23 an overlay published on, updated daily after
24 the mirror updates. With local access to the src files, the script
25 can generate proper Manifest files.
27 The SRC_URIs should point at the (IIRC) archive locations rather than
28 at the src/crontib locations, to ensure that out-of-date overlays still
29 work. And it should have ebuilds for two or three versions, just in
30 case there are any issues.
32 Such an overlay would alleviate most of the issues, and if done at a
33 site (such as osuosl) which is already a cran mirror, would not require
34 excessive network bandwidth.
36 I'd be happy to spend some time a script to build such an overlay, and
37 can do some limited testing of it, but do not have the net bw to run
38 such a script in production, or to fully test it.
40 -JimC
41 --
42 James Cloos <cloos@×××××××.com> OpenPGP: 1024D/ED7DAEA6