[Kde-scm-interest] Data files and the Git move

[Stefan Majewsky]

Hi folks,

I think it should be clear that we need to move to git somewhen, and
it is well known that I think it should happen sooner rather than
later. There are established processes for building split repos, so
the only problem left IMO is the problem of data files.

The matter has been discussed before [1]. The executive summary of the
current situation is:
1. 76% of a kdegames checkout is data files. [2]
2. Git is bad for handling binary files: All history is downloaded
when you clone the repo. The way how GIt compresses the history is
aimed at text files, not at binary files. I think to remember that
someone who used the svn2git rules on the whole kdegames history with
data arrived at a 500MB repository, which is five times the size of
kdelibs.

So the questions which need to be solved before the git transition
are: Where will the data reside? And, if in a separate repository, how
do the dependencies look like?

My answer is: Data shall stay in SVN (in a new module kdegames-data)
for now, and be handled like the kde-wallpapers module after that. I
previously wanted kdegames-data to depend on kdegames, but I see that
the reverse dependency (kdegames depending on kdegames-data) is much
more obvious. Also, it's good for application developers who can use
the distribution's packages for kdegames-data, instead of having to
clone it from the server.

Can we do this?

Greetings
Stefan

[1] http://kde.markmail.org/thread/oxwtkrelvnsywrig
[2]
$ find -type f -regextype posix-egrep -regex
'.*\.(xml|wav|ogg|svg|svgz|jpg|png)' | xargs du -hsc | grep total
86M total
$ du -hsc (*~(.svn|.git)) | grep total
113M total
The latter uses zsh extended glob: (*~(.svn|.git)) matches everything
except .svn or .git