On 8/4/06 8:08 PM, in article [EMAIL PROTECTED], "Reinhard Poetz" <[EMAIL PROTECTED]> wrote:
> Bertrand Delacretaz wrote: > >> Also, if a sync happens while you're uploading stuff, a half-finished release >> will be published. >> >> This looks exactly like the mysterious build problems that Carsten and I were >> chasing at ApacheCon, where he had different versions of some artifacts (from >> another project, not Cocoon) than me in his maven cache, with the same >> version numbers. >> >> I know little about how this all works, but maybe it would make sense to use >> a staging area, outside of the synced directories, to create the release? And >> move it to the "sync zone" when it's ready and checked. >> > yes, that's IMO the way to go. Could some shell script guru set up the > necessary shell scripts for the sync process? What we need is two scripts on > people.apache.org: > > m2repo-to-private.sh [sub-directory]: Using this script you can set back a > particular module, e.g. cocoon-core: "m2repo-to-private.sh > org/apache/cocoon/cocoon-core" > > private-to-m2repo.sh: Using this script we sync our private repo with the > official m2 sync repository. After running this script we can ping > [EMAIL PROTECTED] to do the sync with the ibiblio repo. > > The only thing left is that we should make our private repository available > via http so that we can try out our releases locally. > What about the maven metadata.xml ? Isn't that one updated with each release? We should make sure that there is no risk of breaking something when we start playing with this ourselves. I'll ask for a second opinion on #maven. (but if you ask me, they should be the ones providing a more granular sync service in the first place) Jorg
