Re: Warn recent changes in files being deleted: does it make sense?

Jens-Heiner Rechtien <[email protected]> Wed, 07 Oct 2009 00:15:03 +0200
Newsgroups gmane.comp.openoffice.devel.tools
Message-ID <[email protected]>
Hi,

AFAIK the heuristic works only if the hash over the file is the same. A 
slight change, for instance adapting the name of the file in the license 
header (however stupid such a thing might be) will spoil the heuristic, 
which is why git, like any other sane SCM, supports 'git mv' at least on 
my system. I would be very disappointed if it does not explicitly record 
the rename somewhere (in the index) and is only just the frontend to a 
simple filesystem rename ... :-)

I hate the thought that we now, after all these years, will have finally 
an easy and save way to do renames and people will rely on heuristics to 
achieve that. Not the way to go.

Regards,
    Heiner

PS: The heuristics of hg --addremove and of git are on paar here. This 
is not the place where git gains over hg :-)

Jan Holesovsky wrote:
> Hi Heiner,
> 
> On Tuesday 06 October 2009, Jens-Heiner Rechtien wrote:
> 
>>> Does it really "detect" these cases or does that only work when "hg
>>> rename" was being used. If I would have used only filesystem commands to
>>> rename files would hg recognize that?
>> How should any SCM detect filesystem only renames? This is just asked to
>> much. You have to use "hg rename", of course. Anything else would have
>> to rely on heuristics which might fail, hardly something a SCM can do.
> 
> git does exactly this, ie. has no concept of remembering the renames, and 
> instead relies only on the heuristics.  It works well :-)
> 
> HTH,
> Kendy
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
> 
>