There are various automated processes that commit to my svn (1.5) repo. When I'm branched and merging in from Trunk, merging these commits results in spurious conflicts. Is there any simple way to skip commits by specific users?
+1
A:
I don't think there is. It would be weird too, as it would potentially corrupt the integrity of the source code. Instead, I'd consider two approaches:
- Not doing the automated commits (I haven't run into a situation that required this, but I don't know your usecase well enough)
- Resolve the conflicts automatically by scripting something (always taking one version over another for a specific set of files for example)
Jilles
2009-03-29 16:46:10
(2) probably doable. I know enough about how the automated stuff works to know that integrity issues wouldn't be a problem. If you view merging as something like:svn merge -c a,b,c,.I just want some nifty way to filter on {a,b,c,...}. I guess I can roll my own and then post the solution here.
Jacob
2009-03-29 21:30:13
It's assuming that on one of the two tree's the changes can be thrown away. If both sides are meaningful, (2) will be impossible.
Jilles
2009-03-29 21:36:33
A:
You might consider giving each "automated process" a repo of its own. Unless those processes are doing useful (i.e. creative) work, I wouldn't mix brain-dead changes with brainful ones. I've yet to see an automated process do very much interesting.
jasonnerothin
2009-03-31 04:33:20
A:
This situation calls for a distributed VCS; svn is not one of those.
I'd go for git or bazaar (bzr-pqm is an agent that does just that: integrate incoming merges)
sehe
2009-03-31 21:47:49