Re: Git dies of lack of interest?

2015-12-01 Thread Nick Baker
We at Pentaho made the migration to Git and GitHub from Subversion and
Perforce a few years ago now. At the time we decided to go with multiple
repositories (we're at around 100 right now). In retrospect this was a
large mistake.

Development often spans multiple repositories creating dependent changes
and corresponding pull-requests -a lot of churn on our CI systems.
Branching, tagging, version updates, reformatting are all made more
complicated. Our build teams have scripts to help with this, but it puts a
lot of these tasks out of the range of our single contributors.

We've tried to group together related Projects into the same repository,
but relationships often change between projects. We explored Submodules to
try to provide a unified repository "view", but this was an imperfect
solution.

And there's the confusion from the community and partners about what to
clone, what order to build it in, etc.

Go with a single large repo IMO. It's very easy to view changes affecting
only a particular subdirectory, cutting-down on the history noise.
Similarly the CI SCM plugins are all able to detect which jobs a new commit
affects so a single commit doesn't trigger all jobs to build.

Regards,
-Nick Baker

On Tue, Dec 1, 2015 at 6:41 PM, Richard S. Hall <he...@ungoverned.org>
wrote:

>
> > On Dec 1, 2015, at 17:50, Benson Margulies <bimargul...@gmail.com>
> wrote:
> >
> >
> https://cwiki.apache.org/confluence/display/~bimargul...@gmail.com/Felix+and+Git
> >
> > ?
>
> Seems like a good start, although is that editable by others?
>
> It seems like other technical issues were raised about the approaches, so
> it would be nice to see those added in there by people who have experience.
>
> I admit, for me, SCM is a necessary evil and not something I get too
> exited about. I haven’t seen anything to prefer git over svn or vice versa.
> They’re just different hammers for the same nail.
>
> Still, thinking about the options, it seems like multiple repos creates a
> maintenance headache to some degree. For example, line-ending handling is
> fairly difficult to get configured correctly in git. By having multiple
> repositories, then every repository would have to have this configured
> individually, since stuff like that is good to have configured uniformly.
> Any changes to how we want things uniformly handled would require manual
> propagation of configuration. Of course, this seems like it would be an
> issue in any proliferation of repositories (svn or git).
>
> Or perhaps there is a better way to handle such issues?
>
> -> richard
>
> >
> >
> > On Tue, Dec 1, 2015 at 1:56 PM, Richard S. Hall <he...@ungoverned.org>
> wrote:
> >> On 12/1/15 13:40 , Carsten Ziegeler wrote:
> >>>
> >>> Richard S. Hall wrote
> >>>>
> >>>> Well, the argument to the contrary is perhaps that is makes it more
> >>>> difficult for us as a community to have oversight into releases. It
> >>>> almost assures us that some/many community members will never checkout
> >>>> subprojects that aren't in the repository they normally work. Granted,
> >>>> there is no guarantee of this now, since I can just check out what I
> >>>> want anyway...but at least it is fairly easy for me to do so now and
> it
> >>>> becomes more difficult if everyone spreads to their own repos.
> >>>>
> >>>> So, in that regard, I'm more aligned with Marcel...all or nothing
> makes
> >>>> more sense.
> >>>
> >>> Hmm, ok fair point - however, the *all* is the problematic part where
> we
> >>> couldn't agree on last time (one git repo vs many git repos).
> >>
> >>
> >> But isn't it then incumbent on those wanting such changes to convince
> us one
> >> way or the other?
> >>
> >> Personally, I'd rather just have one big git repo if we are going to
> switch,
> >> if for no other reason than it seems like less overhead. However, I
> admit to
> >> not really knowing the advantages/disadvantages.
> >>
> >> Regardless, at a minimum, perhaps someone should create a documented
> >> pros/cons list for the approaches. This would at least give us a way to
> call
> >> a vote where we can feel somewhat informed about the choices (i.e., stay
> >> with svn, move to one git repo, move to many git repos).
> >>
> >> Better than saying, "there is no consensus, so let's just go our
> separate
> >> ways"...
> >>
> >> -> richard
> >>
> >>
> >>>
> >>> We could still provide a script in the root of svn which checks out the
> >>> moved projects from git and gives the same experience :)
> >>>
> >>> Carsten
> >>
> >>
>
>


Re: Fragment Bundle supplying a Require-Capability seems to not work

2014-03-03 Thread Nick Baker
The fragment was installed first. I know it was working because it was
first simply adding the log4j.xml to the Host classpath, which did work.
Once the fragment was changed to try to add a Capability Requirement, I
expected the host to be Installed, but instead it resolved and started
without the Capability being provided by any bundle.


On Mon, Mar 3, 2014 at 2:12 AM, Marcel Offermans 
marcel.offerm...@luminis.nl wrote:

 Hello Nick,

 From your description I cannot determine if you first installed, resolved
 and started the host and then installed the fragment or if the fragment was
 already installed at the point where the host resolved. I'm asking because
 that's a common pitfall when dealing with fragments: if you install them
 after the host has already been resolved, nothing happens.

 Greetings, Marcel


 On 03 Mar 2014, at 5:42 , Nick Baker codeoncof...@gmail.com wrote:

  Just a heads-up,
 
  According to the specs, it should be possible for a Fragment to add a
  Require-Capability to a Host. I tried this with Log4J to have it wait for
  an Apache FileInstall exploded bundle to supply the Capability, but it
  didn't seem to take. If I get the time, I'll try to gather the work and
  submit a case.
 
  If anyone is familiar with the code, I'd appreciate a quick pointer to
 the
  responsible class. I'd like to submit a patch along with the case.
 
  Thanks,
  Nick Baker




Fragment Bundle supplying a Require-Capability seems to not work

2014-03-02 Thread Nick Baker
Just a heads-up,

According to the specs, it should be possible for a Fragment to add a
Require-Capability to a Host. I tried this with Log4J to have it wait for
an Apache FileInstall exploded bundle to supply the Capability, but it
didn't seem to take. If I get the time, I'll try to gather the work and
submit a case.

If anyone is familiar with the code, I'd appreciate a quick pointer to the
responsible class. I'd like to submit a patch along with the case.

Thanks,
Nick Baker