1 |
On Sunday 10 April 2005 18:39, Aaron Walker wrote: |
2 |
> Regarding GLEP 36[1], solar has asked me to try and figure out a way to |
3 |
> provide both CVS and Subversion for one repository and keep them sync'd |
4 |
> somehow. |
5 |
> |
6 |
> Although this was not the intention of my glep, it would be nice to get |
7 |
> this implemented, as it means all developers of a certain project would not |
8 |
> be forced to use one over the other. |
9 |
> |
10 |
> Unfortunately, I have almost zero experience with repository administration |
11 |
> (CVS or Subversion), so I'm writing this with hopes that some of you guys |
12 |
> have some ideas. |
13 |
> |
14 |
> The only idea I've had is to maybe setup post-commit hooks for both that |
15 |
> commit what was just committed to the other (cvs post-commit commits the |
16 |
> stuff to svn and vice versa). Is this possible? |
17 |
|
18 |
Not really. It is not that hard to go from subversion to cvs. The other way |
19 |
around however will loose information. This is because cvs is not changeset |
20 |
but file based. So if you would like to put sets of changes as one changeset |
21 |
into subversion (in a way that somehow you don't loose information on the |
22 |
identity of the committer) you would need to collect commits, run heuristics |
23 |
on whether they belong together etc. All in all a messy business. |
24 |
|
25 |
Paul |
26 |
|
27 |
-- |
28 |
Paul de Vrieze |
29 |
Gentoo Developer |
30 |
Mail: pauldv@g.o |
31 |
Homepage: http://www.devrieze.net |