On Thu, 05 Jun 2008 10:14:56 +0200
Walter Franzini <[EMAIL PROTECTED]> wrote:

> I'm starting to look at implementing the aegis source backend and I
> see two possibility:
> 
> * use the native replication method (aedist -replay) and then make
>   tailor extract changesets from the local repository.  Looking at
>   tailor source it seems to me the approach used by others source
>   backends.
> 
>   It's not really difficult to do that also for aegis except for the
>   time that may be needed to apply a changeset locally: as an example
>   each change committed in the aegis 4.22 stable branch took not less
>   than 2 hours to be "commited".  The reason for such a long time is
>   due to the need to compile aegis *twice* and run the full regression
>   suite *twice* (development and integration).  With the current
>   development branch the time needed to commit a change-set is
>   measured in minutes (>= 5min)  but not in seconds.
> 
>   I'm not sure a user that want to extract a project from aegis will
>   be happy to wait so much time.

I can imagine very well :-)

I assume there isn't a way to disable the recompilation and the test
suite?

> * the alternative approach is to replicate much of the work of aedist
>   -rec to extract the content of a change-set from an aedist archive
>   skipping the build and test stages.  The aedist archive is a simple
>   bzipped cpio file that contains metadata, complete source of the
>   files and patches.

Well, this would be interesting on it own: several times I wished to
have a source backend for, say, tarballs... We could think about
abstracting it a little, and then adapting to those aegis "bundles".

>   It's a bit more difficult wrt the previous approach because the file
>   containing the metadata need to be parsed, but not too difficult.

Which format is that? Keep in mind that in one way or another, either
when "slurping" the "pending" changesets or step by step as soon as
any single source changeset is applied, tailor needs to assemble its
own ChangeSets objects, enriched with a list of "touched" entries.

> Which one is the way to go (from a tailor user POV)?

I think the second way is a lot better from the tailor *user* POV;
evidently it's a little bit trickier to code, but the advantage is
clear.

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 PROTECTED] |                 -- Fortunato Depero, 1929.
_______________________________________________
Tailor mailing list
Tailor@lists.zooko.com
http://lists.zooko.com/mailman/listinfo/tailor

Reply via email to