Re: [Gnu-arch-users] Moving a sourceforge project to arch

[Miles Bader]

On Thu, Feb 26, 2004 at 01:08:51PM -0800, Michael McCracken wrote:
> I get offers of help from people unfamiliar with version control, and 
> who don't have time to learn CVS. I can tell them how to get the code, 
> but they don't want to commit - they want to send me their changed 
> version. I have to put up, since beggars can't be choosers. Does anyone 
> have an opinion on whether or not arch makes this kind of thing easier 
> to handle? If so, why, and how would you handle it?

It depends.  Arch generally uses a `pull' model for sharing among different
archives: if person CONTR wants to send his changes to person MAINT, CONTR
makes his archive (or a mirror) publicly readable in some manner (e.g.,
stuffs it into an http-accesible tree somewhere), and person MAINT merges
from some branch in it.

In the case where someone can't easily make their archive publicly readable
it's also easy enough to generate standalone changesets in arch.  The
changeset representation is a directory with a bunch of files in it, so the
sender will have to tar up and attach it to his email message using mime or
something, but still it's pretty easy.  [There's been some talk of adding an
`export-changeset' command to tla which would make a readable ascii version
of a changeset that could be easily imported with a corresponding
`import-changeset' command, but nothing's come of it so far]

> Also, because some people are stuck on CVS, I wonder if there's a way to
> use both... Is it worth keeping both arch and cvs in sync? I saw some
> recipes for using both at the same time - is there an easy way, or is that
> the hairy mess I expect?

There's the `tla-cvs-sync' command, in the tla-tools package (which I wrote).
I use it to synchronize my arch emacs branch with the official emacs CVS
tree, and it's pretty easy: one command basically syncs things up in both
directions, with no user intervention necessary for most daily activity
(e.g., it automatically handles new/renamed/deleted files).  If there are
conflicts, of course, you've got to fix them up before preceding.

Here's a wiki page on it:

   http://wiki.gnuarch.org/moin.cgi/tla-cvs-sync

tla-tools is available in my archive:

   address@hidden/tla-tools--devo--0

The archive address@hidden is available from:

   http://sourcecontrol.net/~miles/address@hidden

Some caveats for using tla-cvs-sync:

 * It syncs a single branch at a time -- if you want to keep multiple CVS
   branches in sync with multiple arch branches, you've got to set things up
   for each branch individually

 * It doesn't attempt to intuit `changesets' from CVS (which is a big ball
   of hair in general)

-Miles
-- 
`...the Soviet Union was sliding in to an economic collapse so comprehensive
 that in the end its factories produced not goods but bads: finished products
 less valuable than the raw materials they were made from.'  [The Economist]