Re: [ANN] mtn2git v0.1

[Michael Krelin <hacker@klever.net>]

> On Sat, Sep 13, 2008 at 8:43 PM, Michael Krelin <hacker@klever.net> wrote:
>>> I did look at it, but there are some issues. For starters it seems to
>>> be using extensively the manifest, which would probably make it very
>>> slow. Also, it seems to be keeping the trees of all the revisions.
>> Back when mtn2git script was on the early stages of development and I've had
>> some time I've started doing my own mtn2git gateway in c++ (never finished
>> it, though). For what I remember using manifest was the only way, because
>> the changes information mtn automate returns was incomplete or simply wrong,
>> I don't remember the details... Do check your results carefully if you're
>> trying to go this way. Luckily, being slow due to use of manifest should
>> only matter on initial import.
> 
> It's complicated, but hopefully doable.
> 
> That's why there are two methods; one that does a full checkout of the
> revision and imports it into git, and the other one that uses mtn
> automate and fast-import.
> 
> I'm able to import about 60% of the Pidgin code correctly, but after
> that there are issues. That's the benefit of having two modes... I can
> compare the fast one with the exact one and improve the fast one.

I ended up using some hybrid mode myself, but the point is - there's no 
way to avoid pulling manifests all the time and that doesn't imply 
complete checkout.

Love,
H