On 17 July 2013 15:19, Emmanuel Lécharny <elecha...@gmail.com> wrote: > Le 7/17/13 3:00 PM, sebb a écrit : >> On 17 July 2013 13:26, Emmanuel Lécharny <elecha...@gmail.com> wrote: >>> Le 7/17/13 2:21 PM, sebb a écrit : >>>> On 17 July 2013 10:56, <elecha...@apache.org> wrote: >>>>> Updated Branches: >>>>> refs/heads/trunk 0d04eba45 -> 84b2003fc >>>>> >>>>> >>>>> Revert "[maven-release-plugin] prepare for next development iteration" >>>>> >>>>> This reverts commit 0d04eba4525ca85886c4a03acc20b444ff69a52d. >>>> However trunk is now in an incorrect state, as all the versions are >>>> missing the -SNAPSHOT suffix. >>> The trunk will be fix when the perform will be done. (Actually, it >>> should be the case). >> But meanwhile any builds (including CI builds) will be using the >> non-SNAPSHOT version. >> Not ideal. > > Yep. The plugin should rather create a branch, instead of committing in > trunk...
No, it does not need to do that - at least not with SVN. With SVN the way I do it is: Checkout trunk in a separate clean directory Fix the versions / URLs etc Create the tag directly from the workspace using SVN copy. SVN is clever enough to realise that the only changes are the ones made to the workspace. And at the end of the process, trunk is untouched (so respins are trivial) And you have a nice clean workspace which is available for builds. Once the release succeeds, of course trunk has to be updated to the next SNAPSHOT. But it can never be left in a partially updated state. > In the current case, I was absent around 1h30 between the prepare and > perform, enough for someone to notice the pb. But even if I were to do > the perform immediately, on a bigger project, that would be an issue. > > I'll try to see with the maven fellows if we can improve this process. > > -- > Regards, > Cordialement, > Emmanuel Lécharny > www.iktek.com >
