openoffice-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From florent andré <>
Subject Re: OOO340 to svn
Date Wed, 27 Jul 2011 16:07:38 GMT
Comments inside

On 07/26/2011 05:44 PM, Pedro F. Giffuni wrote:
> --- On Tue, 7/26/11, Christian Lohmaier<>  wrote:
>> Well, do some over-the-thumb-maths please and see that this
>> approach is doomed. 260000 linear revisions (the "easy part"),
>> and you can do 500 in two hours.
>> You need around 1040 hours, or more than 40 days to do the
>> conversion this way.

No ! really ? Thanks to over-the-thumb-maths, my insane brain now have a 
better evaluation of result/time ratio...

> This is certainly a problem :(. Perhaps it will be faster if
> done in the same LAN of the physical repository?

For testing, I set up a machine with svn serveur and the script (so no 
lan delay).

In fact the processing time was almost the same, the costly part of the 
script seems to be the empty folder remove...

> Is the converter that comes with Hg, worse?

Don't work for me. Try with mercurial 1.9 version, but fall down with a 
strange python error.

>> And that is only the time until you very likely hit the
>> point where
>> the conversion will bail out "sorry, can not deal with
>> merges". This
>> is insane...
> It doesn't seem like it will bail out but the script does have
> some documented issues:
> * no handling of branches
> * all commits come from the user running the script

yes true, but the name of the original commiter is in the commit's 
comments, and that's the better things we can have, because not all the 
original commiters have an Apache's svn account, and even if it's the 
case it will really complicated any script.

> * move operations probably won't appear in the history as such
> Nevertheless thanks to Florent for trying and reporting.
> cheers,
> Pedro.

View raw message