Re: Source package names and VCS URLs.
Hi Charles,
On Tue, Apr 24, 2012 at 08:54:49AM +0900, Charles Plessy wrote:
> for some of the packages in our Subversion or Git repositories, the path (URL)
> to the repository does not match its source package name. For instance, epcr
> is in svn://svn.debian.org/debian-med/trunk/packages/ncbi-epcr/. This slightly
> complicates some automated browsing of our repositories.
>
> This could be worked around partly by a symbolic link, or more radically by
> moving the directory. Do you have opinions on that matter ?
Moving is fine for me, specifically in the epcr case we should do it
soonish because I plan to upload the new upstream version that has
incorporated our patch in the next couple of days.
However, in the long run I do not think that your effort to have the
same directory name as the source package name will have success outside
our team (or the teams you can influence). While I consider it
reasonable to normalise the reporitory in this way because it helps
human readers as well (besides automatic tools) there will be always
reasons to name the repository differently. So finally the tools
probably need to cope with situations where the repository name is
simply something else than the source package name. We do even have
this here in the case of substructures ('R', 'rostlab', now new 'soap')
and IMHO it is better to find automatic ways to detect this structure
rather than specifying a set of exceptions.
BTW, the explicit specification of such subdirectories in
scripts/umegaya-guess-url looks suspicious to me. You will never cope
with all those different ways all the Debian teams might invent to
structure their repository. I have no idea how this code is used in
umegaya but when I stumbled upon it I wanted to express my doubt that
this could become a general solution. IMHO we need to find a way to
extract some kind of tree inside a SVN and seek for "trunk/debian" dirs
(no idea if this is sufficient) instead of trusting a flat layout with
some exceptions of subdirectories.
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: