Re: [Gnu-arch-users] Using arch for offline development

[Wazow]

"Stephen J. Turnbull" <address@hidden> writes:
> The only thing that worries me about the local archive, backup mirror
> strategy is that I don't see how he automates the push-mirror any more
> effectively than under the current situation with CVS.  It's still
> something that needs to happen when the network comes up, and if
> that's typically at the beginning of his session, he's still going to
> have issues with resource usage at an inconvenient time.

This worries me  too, but it is indeed slightly  more flexible than with
CVS.  With  CVS I am  stuck to have  either local or  remote repository.
Using a  remote repository  with CVS  does not allow  me to  commit when
offline.  In  principle I  could hack an  own version  of offline-commit
wchich would just  collect patches and log entries  and then replay them
when going online.  But this would be nearly a reimplementation of archs
branching using CVS back-end  (well ... sort-of). Using local repository
with CVS works well, but it is unsafe. I cannot push the hole repository
out of  the machine whenever being  offline (bcs this is  a huge tarball
with the entire cvs tree), so the mirror is pretty much out of sync most
of the  time. In principle  I can loose  a day long  work, even if  I am
working online.   Finally I have no  repository on the net  that I could
access in  case I do not  have my laptop with  me, but I  have access to
some  other  machine.   This   mirror  swapping  with  tla  (or  perhaps
mirror-pushing  as someone  suggested) gives  me safer  &  more flexible
online working conditions,  while it has all the  advantages of having a
local CVS repository when offline.

But while I am thinking on it, I start to agree that instead of hacking
mirrors it would be better to use branches from the remote tree as this
allows me to give others access to the repository. Also it allows me to
use it remotely from other machines than my laptop if need.  So what I
would really like is a little script (which I can write myself, with a
little hint from you on what tla mechanisms I shall use) that would
commit my notebook-branch changes to my server-branch changes, but not
in one shot (all notebook changesets as one server changeset) but
consecutively one by one, so that server log entries are the same as
notebook log entries (and the changesets are really the same, unless
someone interphered on the way). Then I could just use the remote
archive when online, update my notebook-branch when going offline, use
it for coding offline and finally commit to server-branch using the
magic script, when going online again. 

I start to think that would be cleaner, more tla-spirit and more
flexible in the end. But what should this script use?

Andrzej