Re: Alioth: html output dir linked to vasks???
On Thu, Oct 13, 2011 at 12:39:55AM +0200, David Paleino wrote:
> (no need to CC me, I never unsubscribed from debian-med@, eh :))
:-)
> > I was advised by Alioth admins to do so and they installed all
> > preconditions to run the tools that are creating the tasks pages.
>
> Uhm, ok.
> I knew about the symlink just yesterday -- there are people using this
> method for other projects (Ansgar Burchardt for PET, for example), so I thought
> it was the "accepted way" of doing things.
Well, for PET only purpose this method is probably perfectly OK. But we
are doing more stuff than only PET - thus we experience this conflict.
> Could you expand this point?
> What does this "publish" target do? Does it scp things from vasks to wagner?
See
http://anonscm.debian.org/viewvc/blends/blends/trunk/blends/doc/Makefile?view=markup
IMHO this manual triggering of publishing something has some advantages
because it enables you to commit some intermediate status which is not
automatically subject for publishing. I personally would prefer this
anyway so I do not really see a constraint in the fact that postcommit
hooks do not work any more.
> I'm not entirely sure that's the most correct way. Those are the same admins we
> made the symlink wagner:/srv/home pointing to vasks :)
>
> Anyway, we can contact them, asking for clarification.
However, I would like to get the tasks pages working *before* we have
sorted out this. So I would love to revert your change right now
(risking that postcommit hooks will not work for the moment - PET should
keep on working if you copy the code manually ... at least I think so).
> > No problem. I do see two ways to get things working again:
> >
> > 1. Revert the general symlink for htdocs and symlink only to those
> > files / directories which really profit from this step.
>
> 18 symlinks seem too much to me :)
>
> dapal@wagner:~/debian-med/htdocs$ ls |wc -l
> 19
>
> (-1 → tasks/ directory, which wouldn't be symlinked)
Even if it is actually 17 (-1 for bugs/) I agree that this is hardly a
good idea because when creating a new file this will break.
> > 2. Revert the symlink and use a publish target as I suggested (and
> > do not use the postcommit hook to move things around
>
> This is what I did not understand :)
Please see the Makefile linked above. Currently your workflow is like
1. Changing something
2. svn commit --> changes are populated
My proposed workflow is:
1. Changing something
2. svn commit (NOTHING will be populated)
3. make publish --> changes are populated
While this is actually one step more I *really* think that this one more
step is reasonable. Thinking once more before publishing is IMHO a sane
thing to do and I was actually never really happy about the shorter
workflow.
I'm somehow tempted to revert the symling soonish because I need working
tasks pages (nowish) because I need to demonstrate something here. For
the moment nothing should be broken by this change back because the
working code is just populated, right?
BTW, did you created a copy of the old htdocs dir before creating the
symlink?
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: