On Feb 17, 2010, at 08:36, dcz wrote: > Le 17/02/2010 15:24, Andy Levy a écrit : >> > >> There is no "dry run" for a commit. It's conceivable that you could >> check the user ID in the pre-commit, and if the committer is JD, >> reject the commit while mailing a diff to SD which represents the >> change being attempted. This would require a lot of overhead on the >> server though. >> >> Honestly, I think you're trying to apply an excessively technical >> solution to a fundamentally non-technical idea. >> > > You're probably right, I was just thinking that a bit more automation could > be handy. > Though solution 1) is very close to do it, could use a mailing list to warn > all SD at a time and see who handles first. > I would just have liked to skip the JD "putting together a patch" step and > have it done by the server, but ho well.
I think it would be easy -- for the server or the JD manually -- to put together a patch -- it's simply: svn di > patchfile It could even be mailed to the SD straightaway: svn di | mail s...@example.com You could try to build this into the pre-commit script, but I'd think the JD would start getting annoyed that his patch review process keeps giving him negative feedback by telling him "Your commit was rejected with the following message" (or whatever the verbiage is that Subversion prints out). I'd also say it's confusing that the JD has to run "svn commit" when he's not really committing anything to the repository. Instead I'd suggest you write a new script, call it svnreview.sh or something, and have the JD run that instead. This script runs "svn di", maybe some other sanity checks first, then sends the email.
