SV: [PATCH 0 of 1] convert: make convert from cvs recognise branchpoints correctly

[Henrik Stuart]

> From: gerg.ward at gmail.com [mailto:gerg.ward at gmail.com] on behalf of 
> Greg Ward
> Sent: 8. april 2009 03:00
> To: Henrik Stuart
> Cc: mercurial-devel at selenic.com
> Subject: Re: [PATCH 0 of 1] convert: make convert from cvs recognise
> branch points correctly
> 
> On Tue, Apr 7, 2009 at 7:15 AM, Henrik Stuart <henrik.stuart at edlund.dk>
> wrote:
> > This patch fixes issue 1447.
> 
> Two comments: MPM griped the other day that "[PATCH 0]" emails often
> contain detailed info that belongs in the commit message.  *I* think
> your commit message is fine, but be careful you don't annoy Matt.  You
> want him to accept your patch.  ;-)

That's why I added the extra information in the patch description. The 
PATCH 0 description is mostly just a recap of issue 1447, but I'm not
against adding further information to the commit message if need be.

> Also: adding another attr to logentry will break incremental
> conversions because of old pickles in the cache.  We both made this
> mistake recently when adding 'synthetic' and 'mergepoint'!  I see a
> patch in crew that works around it by using getattr(e, 'synthetic')
> instead of e.synthetic.  Not the prettiest, but it should work.  You
> should probably do similarly.

Right you are. Do note that almost all existing CVS conversions that
contain branches where the initial commit doesn't touch /all/ relevant
files will be wrong (my feeling is that this will be vast majority). 
I'll probably prefer to detect this situation and issue a warning. 

Revised patch will be forthcoming shortly. Thanks.

-- 
Kind regards,
  Henrik