Converting HUGE (12GB) SVN repo to Hg

[Rock Hymas]

As Dirkjan mentioned, hgsubversion might be a better way to go, given what you're trying to do. You can also configure it to ignore branches, which should speed things up too, using the branchmap option on clone.

-----Original Message-----
From: Philip Pemberton [mailto:lists at philpem.me.uk] 
Sent: Thursday, April 22, 2010 8:39 AM
To: Rock Hymas
Cc: mercurial at selenic.com
Subject: Re: Converting HUGE (12GB) SVN repo to Hg

Rock Hymas wrote:
> You did the right thing by converting from a svnsync'd repository.

Oh good, at least I did one thing right! :)

> However, in all likelihood the conversion won't complete, hg will run
> out of memory first. You can rerun it over and over and that might
> work as long as there aren't any very large revisions.

:-(

> Is your repo laid out using trunk/branches/tags at the root, or with
> project directories that each have trunk/branches/tags? If the
> latter, then you most certainly want to convert each project
> separately.

As far as I can tell, it's just one REALLY BIG project with a lot of 
history:
   http://gcc.gnu.org/viewcvs/

It follows the normal branches/tags/trunk structure, and ideally I'd 
like to keep the tags. What I want to do at the end of this is use 'hg 
bisect' to track down a very pervasive bug that's breaking Linux kernel 
builds on the lm32 platform...

> Are there portions of the repo that are purely binary and/or
> compressed files?

I don't think so.. AIUI it's all text files (source code) and huge shell 
scripts (automake/autoconf fluff).

Thanks,
-- 
Phil.
lists at philpem.me.uk
http://www.philpem.me.uk/