Re: [Monotone-devel] cvsimport branch reconstruction

[Markus Schiltknecht]

Hi,

Uff, I'm having a hard time getting used to C++. I thought I may 
probably just ask here. When exactly do I use shared_ptr?

I want to have a map for
(author_id, changelog_id) -> vector<resync_information>

resync_information holds three timestamps and yet another vector of 
cvs_events.

Comming from C, I'm thinking in terms of pointers, so I'm currently 
trying this:

map<metadata, vector< shared_ptr<resync_information> > > resync_revs;

Under what circumstances can I leave away the shared_ptr? Is that the 
right way to build such a map?

Maybe I'm just to tired. I'll go to bed now.

Best wishes

Markus


Markus Schiltknecht wrote:
> Hi,
>
> I've just fixed some outstanding issues with the 
> cvsimport-branch-reconstruction branch. Now all additional tests for cvs 
> imports succeed. It's still my old 'home-brewn' improvement of the 
> algorithm (taking only time into account).
>
> I have also added the 'resync' step from cvs2svn. This tries to fix 
> erroneous timestamps within one single RCS file. Such nasty things 
> should only occur due to clock skew on the CVS server, though.
>
> I'll try to improve cvsimport further so that it resembles the cvs2svn 
> algorithm even more. And the wiki just got to know about it:
> http://www.venge.net/monotone/wiki/BranchStatuses
>
> Regards
>
> Markus
>
>
> _______________________________________________
> Monotone-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/monotone-devel