Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [jgit-dev] Re: Copyright/License Template for JGit

Robin Rosenberg <robin.rosenberg@xxxxxxxxxx> wrote:
> fredagen den 19 februari 2010 12.43.45 skrev  Matthias Sohn:
> > 2010/2/17 Shawn O. Pearce <spearce@xxxxxxxxxxx>
> > 
> > > Due to the history rewrite I'll also have to delete all current and
> > > historical *JGit* changes out of our Gerrit Code Review database.
> > > EGit changes will not be affected.  In theory I could try to patch
> > > up the database onto the rewritten history, but line numbers will get
> > > all screwed up due to the header change, so its just not worth it.
> > 
> > Could we save the hitorical JGit reviews by regenerating all review mails
> > (those Gerrit sends if you register to watch projects) into yet another
> > mailing list "jgit-review" ?
> This don't contain the references code. Can the old ones be retained, 
> referencing the old commits? For the unapplied patches we could use
> the mail trail to reconstruct the most important comments. Gah, what
> a mess, Shawn, do what you think is most reasonable. 

Right.  It is a mess.

What I can do is dump the database and store it somewhere.  Its not
that big.  I probably could archive it in some side branch on or something, along with the pre-rewrite commits just
in case someone really cares about matching stuff up.

In theory I could fix the database for accepted changes by letting
filter-branch rewrite the refs/changes/ namespace and use that to
update the database after the rewrite is done.  I haven't tested a
filter-branch rewrite of hundreds of refs like that.  I'd hope it
does the right thing in that case.  :-)

For abandoned changes, I might be able to do the same thing.
Or we probably just have to bite the bullet and throw them away.
I could try to archive them on outside of the foundation's
reach, or on  But there won't be any review comments.
This is why I really want to work on git notes support in JGit in
the coming weeks, so I can move the Gerrit review database data
into git notes and make it distributed.

Regarding pending changes that haven't yet merged, I'll have to
throw them out too, and the authors will need to rebase on top of
the rewritten history and upload all over again.  That does mean
the pending comments will be lost.  Fortnately there are only a
handful of things pending, most of it is mine.  Unfortunately,
there are a lot of useful comments on those commits which I wanted
to address before the code submitted.

I'll see what I can do with a reasonable amount of time.  I don't
want to spend more than 2 days on the rewrite.  I'd rather be
writing useful code.

> This story probably deserved a blog entry somewhere for others to learn
> from.

Probably.  But apparently I only blog in anger.


Back to the top