Converting CVS repository with huge binary files

[Colin Caughie]

> -----Original Message-----
> From: mercurial-bounces at selenic.com [mailto:mercurial-
> bounces at selenic.com] On Behalf Of Arne Babenhauserheide
> Sent: 10 June 2009 21:21
> To: mercurial at selenic.com
> Cc: Stephen Rasku
> Subject: Re: Converting CVS repository with huge binary files
>
> Am Mittwoch, 10. Juni 2009 17:38:05 schrieb Greg Ward:
> > On Wed, Jun 10, 2009 at 11:03 AM, Stephen Rasku<srasku at gmail.com>
> wrote:
> > > I just tried to convert a CVS module to Mercurial.  It has huge
> > > binary files and it failed after 6 hours after converting only
> 10 of
> > > over 60 changesets.  I was using the --datesort option.  Would
> it
> > > help to omit it?  Is there anything else I can do?  It looks
> like
> > > the CVS server has 4G of RAM.
>
> You can try using cvs2svn to first get the cvs into a nicer shape
> where commits aren't tied to files. From there Mercurial can convert
> much easier.

Another thing you might try is first converting to Mercurial without the --datesort option, and then converting again (from hg to hg) with --datesort.

That's what we did with our Synergy/CM repositories (using a hand rolled conversion tool to do the first stage) and it worked pretty well. The repo size after the datesort conversion wasn't much bigger than before it.

Having said that we did exclude all of our big binary files from the conversion; we're looking for a non-Mercurial means to control those.

Cheers,
Colin


Latest News at: http://www.indigovision.com/news2009.php