[VTK ARB] some git info[Fwd: Re: [Nipy-devel] [nipype] 0.1 tagged]

Mon Oct 19 18:49:56 EDT 2009

Hi All,

I use both Windows and Linux. For Windows I am using msysgit and it seems
Ok.

I have been experimenting with the svn+git solution here and it seems Ok.
One thing I sometimes forget is to do a "git svn rebase" but git seems to be
quite forgiving! Oddly enough I find the command line interface for git in
windows more useful than TortiseGit but on the other hand TortiseGit gives a
good visual indication it the state of the repository. Maybe when I am more
confident I will use TortiseGit.

One thing ... even on a small repository the initial checkout from the svn
can take a long time. That is using the command "git svn clone". Oddly
enough it is really slow in windows but much faster in Linux. I also got
some Perl errors in Windows and had to do a fetch and a rebase to fix the
local repository. I suspect these errors were caused by the network
connection. However git is robust!

Another issue I haven't solved is that Git Bash always asks for the password
when accessing the svn repository - maybe I need to set something but I
haven't found it yet. No such issues in Linux.

Git takes a little getting used to but overall it seems really robust. I
like the fact you have a full history on your machine, you can try things
and then revert quite easily.

I have started to use it for ad-hoc developments. It is really nice to have
a repository with a complete history on your machine for this sort of work.

I guess my advice to windows users is to first get familiar with the command
line interface: Git Bash and try out the gui: Git GUI. Git GUI seems
reasonable to use. The help for git is quite good.

Regards
   Andrew

-----Original Message-----
From: arb-bounces at vtk.org [mailto:arb-bounces at vtk.org] On Behalf Of Steve
Pieper
Sent: Monday, 19 October 2009 23:50
To: arb at vtk.org
Subject: [VTK ARB] some git info[Fwd: Re: [Nipy-devel] [nipype] 0.1 tagged]

Hi -

FYI there's some more git info below for the curious.

 From what I've been able to discern, I think the svn+git solution Paolo 
suggested sounds like a good solution.

-Steve

-------- Original Message --------
Subject: Re: [Nipy-devel] [nipype] 0.1 tagged
Date: Sun, 18 Oct 2009 15:07:52 -0700
From: Fernando Perez <fperez.net at gmail.com>
Reply-To: nipy-devel at neuroimaging.scipy.org
To: nipy-devel <nipy-devel at neuroimaging.scipy.org>
References: 
<764e38540910051921i3f41259en14978bc149389e66 at mail.gmail.com> 
<20091006132051.GY17964 at onerussian.com> 
<764e38540910061040y78bf9afclffc8de5b748a33c1 at mail.gmail.com> 
<db6b5ecc0910151424l473cb630vaa3305dea0c38ae3 at mail.gmail.com> 
<4AD861B7.8040802 at bwh.harvard.edu>

Hi Steve,

2009/10/16 Steve Pieper <pieper at bwh.harvard.edu>:
> Hi Fernando -
>
> A couple projects I'm involved with are considering git and I wonder if
> you could help with a couple questions.  (Since you are now the
> git-go-to-guy. :)

Well, I only have  limited experience with it so far, so I'm still
very much a newbie.  But I'm happy to share what little I know :)

> Can that 'history folding' issue come up in git or was that a bzr
> specific issue?

I don't have a 100% certain answer to this, but Ariel and I looked
into it by explicitly trying to force it to happen, and it didn't.
The reason I think it doesn't happen is because git does not try to
keep actual revision numbers it wants to keep on increasing, but only
commit hashes.  So if two copies of a given repo evolve differently
and then there are pushes from both to a shared copy, git can still
handle this because it simply makes each merge operation and records
the resulting commit hash, and as long as there are no conflicts,
there's no problem.

So it *may* be that a similar problem can still occur with git, but so
far we haven't seen it even while deliberately trying to cause it.
Git's merge algorithms are ridiculously robust, it seems so far.

> Also, I've heard that git on windows is not as mature.  Is this just a
> matter of needing to install the right client, or might people run into
> mysterious problems if they try to use windows?

I don't use windows, but what I've heard is that the newer msysgit client:

http://code.google.com/p/msysgit/

is actually not that bad.  However, the hands-on opinion of a windows
user should count for a lot more than mine.  One issue with Windows
may be that I think the *gui* git clients aren't very mature (not at
the level of TortoiseSVN at least).  Since Windows users are typically
less comfortable with command-line environments than *nix ones, this
may be a factor to consider.

Having said that, these two guides I just found make it look like it's
a fairly reasonable setup:

http://www.lostechies.com/blogs/jason_meridth/archive/2009/06/01/git-for-win
dows-developers-git-series-part-1.aspx

http://nathanj.github.com/gitguide/tour.html

I hope this helps...

Cheers,

f

ps - Steve, are you here at SfN?  If so drop me a line so we can try
to meet up...
_______________________________________________
Nipy-devel mailing list
Nipy-devel at neuroimaging.scipy.org
http://mail.scipy.org/mailman/listinfo/nipy-devel
_______________________________________________
Arb mailing list
Arb at vtk.org
http://public.kitware.com/cgi-bin/mailman/listinfo/arb