Re: Helping new installs find the right dependencies?

[Michael Richardson <..hidden..>]

Erik Huelsmann <..hidden..> wrote:
    > How do others feel about providing such a deps-tarball? Would it help
    > solve the dependencies issue? Does anybody have experience doing so for
    > a Perl project (I have experience for a C project, but that's a
    > different matter entirely)?

a) have the set of dependancies available (individual tarballs) on www.ledgersmb.org.
b) have a shell script that runs cpan with the right arguments to pull from
   above.  It seems that install from URL isn't something cpan does directly,
   perhaps:
     http://blogs.perl.org/users/lestrrat/2013/03/the-main-problem-with-cpan-modules-on-github.html

   cpanfile will work.
   Aha: cpanp
      http://gugod.org/2010/02/tip-install-a-cpan-module-directly-from-url/

>known-working versions of our dependencies, the long-term strategy would be to
>try and get in as many distributions as we can.

The problem is that ledgersmb really moves too fast for most of the distros,
and once you install the distro version, it can be difficult to
replace/upgrade it.

My suggestion has been to have a "ledgersmb-runtime" virtual package which
references all the dependancies.  The distro "ledgersmb" package depends upon
it, and provides ledgersmb, but if you need something newer, then you can
install everything from ledgersmb-runtime.

I suggest that ledgersmb-runtime could be a package that we include in the
ledgersmb tree itself.  For Debian, I am in the process of bringing a machine
up to date, and I could produce that by the end of April.  If I can remember
how to build .spec files, I might be persuaded to do one for Fedora.

--
]               Never tell me the odds!                 | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works        | network architect  [
]     ..hidden..  http://www.sandelman.ca/        |   ruby on rails    [