[FFmpeg-devel] [CHECKLIST] what is preventing us from switch to git

[Stefano Sabatini]

On date Saturday 2010-06-12 12:15:28 +0200, Michael Niedermayer encoded:
> On Sat, Jun 12, 2010 at 10:37:17AM +0200, Luca Barbato wrote:
> > I noticed that more and more people started to use git instead of svn to 
> > contribute to ffmpeg, I'm wondering if the times are now mature to switch 
> > to git.
> >
> > Could you please raise (again) your concerns about switching to git?
> 
> personally i dont really care if we use svn,cvs or git the differences
> dont matter much in normal everyday development if one has a write account.

Relevant thread:
http://thread.gmane.org/gmane.comp.video.ffmpeg.devel/88150

The last time it was discussed we seemed to agree on the migration,
what we lacked was someone actually doing the thing.

One of the advantage of using git would be to have FFmpeg+libswscale
in a unique repo, currently making the checkout of an old version is
painful.

Also we may have some commits which require changes to both libswscale
and FFmpeg, currently this cannot be done with a single atomic commit,
same for having a git branch with committs applied to both repos.

Also even using git-svn some commit needs to be done using SVN, for
example a file renaming.

> a few points id like to raise though are that
> * a switch if it is done should not
>   be done during SOC to not burden the students unneccessarily
> * It should be announced at least 2 weeks before its done so people
>   have a bit of time to get used to git and maybe to commit outstanding
>   changes from their svn checkout
> * We need to document, a short cheat sheet that lists the
>   git equivalents to the common svn commands (that of course before
>   a switch is done)

http://git-scm.com/course/svn.html
http://wiki.videolan.org/Git
(many others, Google is your Master...)

No need to add another document.

>   This also reminds me that quite a bit of code in ffmpeg is badly lacking
>   documentation, one i ran into a few times was for example the scripts and
>   how to make sure thir cross compilation doesnt break. (A self test during
>   regression testing that tests them more thorughout might also be usefull)

I cannot but agree, currently my patch on the libavfilter test is
frozen exactly because of that.

Regards.
-- 
FFmpeg = Free Fabulous Muttering Pure Efficient Game