Re: tailor monotone update

I'm CCing the revctrl ML where other people may be contribute.

>>>>> "Markus" == Markus Schiltknecht 
>>>>> <markus-XHb0b+agOrdyDzI6CaY1VQ@xxxxxxxxxxxxxxxx> writes:

    Markus> So I've added the options --monotone_db and
    Markus> --monotone_branch. These get passed down to
    Markus> _initializeWorkingDir() calls, so I've had to patch those
    Markus> calls and function definitions. I did simply add 'options'
    Markus> as an argument to _initializeWorkingDir(). That way, we
    Markus> could also reduce the other arguments, since they are
    Markus> included in options. But I didn't dare to do that,
    Markus> yet. You know python and your own code better: what's the
    Markus> right way to go?

I'm not so happy of introducing backend specific options like these,
where other backends could make use of semantically similar
information. Better names could be target-repository and
target-module, then each backend may decide on it own how to map
them. 

So, I'd prefer to make it somewhat more general, maybe splitting
_initializeWorkingDir() into two steps, with the first that eventually
does the setup, and the latter that actually performs the import.

But yes, your idea of reducing arguments using just options is the way
to go.

    Markus> For monotone to work properly at bootstrapping, I had to
    Markus> change the first commit to work within the monotone
    Markus> work-directory. That's in
    Markus> target.py:initializeNewWorkingDir(). Instead of committing
    Markus> the subdir I do a commit within subdir without any
    Markus> arguments. I tested this with only monotone (which doesn't
    Markus> work the other way around). I expect CVS and subversion to
    Markus> also work like that, but I didn't test.

Uhm, it may cause problem with other backend, when the work-dir isn't
yet committed to the repository... Can't this be done only for
monotone if needed and if this does not apply to it?

    Markus> Unfortunately tailor still aborts when importing from CVS
    Markus> into tailor as soon as there are deleted directories. I
    Markus> think monotone should be updated to handle directories
    Markus> better in general, but I will try to circumvent the
    Markus> problem in tailor.

Can you be more specific? What's the problem?

    Markus> I've taken a look at the tests. As far as I understand,
    Markus> you are only testing upstream functionality, not any VCS
    Markus> as target.

Yes :-( Evenif the current tests more or less cover what I consider
the most difficult part, I never had the right idea on how to
implement the other side tests.

    Markus> I thought about implementing a test for
    Markus> monotone as target, but... where to put all the files? how
    Markus> do I check for correctness?... still many questions
    Markus> open. Is there a similar test I could copy and adjust for
    Markus> monotone?

Maybe we should start setting up a reference repository, and migrate
that to the various backends, then going from there in the other way
around and see if we get the same result.

    Markus> When using monotone as upstream VCS.. how do I cope with
    Markus> multiple heads?  Converting from CVS to monotone is easy,
    Markus> because monotone has all the functionality CVS has (or at
    Markus> least concerning the functionality required for
    Markus> tailor). But not the other way around. Does it make sense
    Markus> at all to convert a monotone repository to another one
    Markus> (considering you will most probably loose some
    Markus> information)? For an example that's not trivial to convert
    Markus> to CVS or subversion, please check the current
    Markus> net.venge.monotone branch ancestry graph at:
    Markus> 
http://grahame.angrygoats.net/viewmtn/revision.psp?id=ebc417b19960d8e2708ff67b068a3fbd3c1a98e0&ancestry_limit=20

This is a question I really would like to hear somebody else idea.
While I think that may quickly go beyond tailor's business, as it
currently deal with a single "branch" at a time, this would probably
require reworking the source backend to understand things like
"tagging" and "branching", and the target backend to react
appropriately, when it can.

    Markus> P.S.: sorry for sending only a plain patch. I just didn't
    Markus> have the curage yet to read a tutorial on yet another VCS
    Markus> (darcs)...

I will apply some bits of it, but for the rest I'll need to have some
more time.

    Markus> P.P.S.: no mailing list, yet? It would be great to have an
    Markus> archive of such discussions for reference.

See http://lists.zooko.com/mailman/listinfo/revctrl

ciao, lele.
-- 
nickname: Lele Gaifax   | Quando vivrò di quello che ho pensato ieri
real: Emanuele Gaifas   | comincerò ad aver paura di chi mi copia.
email: lele-Fk80lsQIhWConA0d6jMUrA@xxxxxxxxxxxxxxxx     |               -- 
Fortunato Depero, 1929.