Re: Source control backends

[Ali Afshar <aafshar gmail com>]

Stephen Kennedy wrote:
> The SCM backend interface is showing its age and its humble CVS
> origins. It breaks regularly and theres no unit tests. There's no
> interface to history.
>
> In short it needs a major overhaul or replacement.
>
> Does anybody know of a possible replacement? Some time ago the pida
> project borrowed the vc library and rewrote it (now called anyvc).
> Any others we should consider
Hi everyone,

I have only made brief contributions to the current incarnation of 
anyvc, but I can take a stab at it. I am sure Ronny (the main author) 
will correct me on anything or expand on any queries.

Most importantly (imo) the new anyvc is a totally separate package, and 
is published on the Package Index as "anyvc", so it would be perfect for 
Meld to use and remain independent of the back-end. We did this 
primarily to facilitate collaboration in the community. A situation of 
"write one set of integration plugins conforming to a single interface" 
which could then be reused throughout a very broad spectrum of 
applications: [IDEs, Diff/merge tools,  Online code browsers, universal 
command line libraries, ...] would be a very useful situation.

It would very much encourage users and developers of 
source-control-system software to add their own integration to an 
excellent quality.

So what do we have?

   - Working directory support for some systems. That means: "all 
commands that affect the working directory, file listings updates etc. 
(A few of the less-known systems are missing)
   - Unit tests


Ronny is working on a complete feature matrix at: 
http://pida.co.uk/trac/wiki/AnyvcStatus

What do we want?

- History support for all the systems: viewing the history graph, 
getting metadata/diffs.

What do we need? Well, we can get pretty far with most version control 
systems. We use Mercurial mostly, so obviously our mercurial support is 
the best, and obviously other systems might lack love and features. So 
join in, please.



Ali