[webkit-dev] An incremental approach (was Re: UPDATED Re: Version control survey)

[Tor Arne Vestbø]

On 11.03.12 00:08, Alp Toker wrote:
> The way I see it, a better mirror would address:
 >
> *Author Names
>
> *The committer names don't have the full author name in the git mirror
> right now, just an SVN id. This info could be extracted out of a
> database or the ChangeLog message if one exists, during import. People
> switch companies and email addresses over time, so that would have to be
> accounted for.

The mirrors at http://gitorious.org/webkit/webkit and 
https://github.com/webkit/webkit (same repo), use an author-script 
during the import to resolve author names based on the committers list. 
The script could be more intelligent and pick up author names from 
"Patch by" in the commit message or the changelog itself.

> *Layout and repo size
> *
> The git repository with full history is enormous.
>
> A proposal (or even better, proof of concept) for git repository layout
> where the 'heavy' generated paths are split out into git submodules
> separate from the source code would make me feel more comfortable with
> the whole idea. Also, should be possible to do a shallow clone of these
> yet still be able to commit and push back upstream (if git supports
> this, git experts?)

I've done some prototyping on how to do a smaller mirror, using git 
submodules or tools such as git-annex or git-media. So far the issue is 
that if you want to commit to SVN using git-svn none of these techniques 
can be used, which makes the smaller mirror less useful.

There was a thread on the git mailing list that mentioned the possibly 
of writing a git-fast-import/export backend to solve this, ie to lazily 
populate the layout-test results, but I haven't had time to look into 
that further.

So the conclusion so far is that it's not feasible to keep an 
incremental SVN-mirror that does on-the-fly pruning of layout-test 
results (into submodules or similar) while still being usable with 
git-svn. Ideas welcome!

tor arne