[darcs-devel] Support for darcshive repositories?

[David Roundy]

On Sun, Dec 05, 2004 at 11:32:08PM +1000, Anthony Towns wrote:
> David Roundy wrote:
> >And perhaps adding a description of DARCS_APPLY_FOO to Apply.lhs.
> 
> YM Push.lhs? I couldn't see anything obvious to say about that anywhere...

Right.  Well, roughly to tell users that if they want to push to a
repository available over some other protocol than ssh/local, they'll have
to define a DARCS_APPLY_PROTOCOL.

Also, I'd lean towards explaining in the docs on DARCS_APPLY_FOO that in
general, this command should simply run darcs apply on the remote
repository.  That's not what darcshive does, but if one wanted to create a
cgi script or something that allowed you to push over http, that's what
you'd have to do.

> The attached patch set includes some documentation, an implementation 
> for DARCS_MGET_*, and an implementation for DARCS_MGETMAX.
> 
> MGET's really kind-of handy:

:) Applied.

The one thing that bothers me a bit about the code is that we keep checking
the environment variable DARCS_GET_FOO again and again, if MGET isn't
defined.  It might be nice to break out of copyRemote a "getRemote"
function that accepts the "get" output of maybeURLCmd, so we wouldn't need
to keep rerunning maybeURLCmd.  Probably it doesn't make a difference, but
it would be prettier.

> Next up is making darcshive clueful enough to be able to store darcs.
> Unfortunately that means dealing with all sorts of special case
> behaviour, like old-style timestamps :(

Alas, that is indeed some ugly code.  That and the old merger 0.9s.  I've
been meaning some day to introduce an option to optimize that gets rid of
merger 0.9s... although I guess that since you don't need to interperet the
patches, the merver 0.9s don't bother darcshive.  Still, it'd be nice to
get rid of them...
-- 
David Roundy
http://www.darcs.net