On 20 August 2014 10:04, Robbie Gemmell <[email protected]> wrote:
> > On 19 August 2014 22:47, Justin Ross <[email protected]> wrote: > >> >> Here's a preview of what I've got going. This example run is against >> trunk, but the same structure would be reflected into 0.30. Note that the >> dispatch and proton artifacts get removed before doing the "Qpid" release. >> >> http://people.apache.org/~jross/quirk/20140819/ >> http://people.apache.org/~jross/quirk/20140819/java-binaries/ >> >> I added qpid-qmf2, Fraser's stuff; I'm thinking about renaming the archive >> to qpid-java-qmf-tools. > > > Yeah, its a bit annoying to name, and not aided by the existing artifact > names. > > >> Correspondingly, the python tools archive is back >> to being just python code >> >> The java binaries are now precisely the same as "mvn package" generates. >> > > Great > > >> These include websocket and perftests binaries that no one has yet >> expressed a desire for. >> > > If we distribute the client websocket plugin, we should update the broker > to depend on it so its binary includes the matching broker plugin. I'll do > that now. > > I'd be fine without the perftests binary being included. We have never > distributed it in the past, and the maven build is currently set not to > deploy any of the output for that module as a result, which we should > perhaps change if we want to begin distributing it on the site. It is only > being built because the old Ant build system had an optional way to do so, > and in the end I tied it to the package phase for consistency with the > other modules. I did consider leaving it requiring a more typical explicit > "mvn assembly:single" invocation to create that archive, saving the > precious ~0.x seconds and additional storage during the build process ;) > I also meant to mention the .revision files. Are we going to distribute all of those at the end? I know we had a '.svnversion' file in the past, though having a quick peek I dont see any other Apache projects which do that, but having one of them per archive makes things seem a bit cluttered somehow. That said, the same quick peek also made me realise everyone else does per-archive .md5 and .sha1 files which we never have, instead having a single SHA1SUM file containing all the hashes. That file seems to have dissappeared, which we should probably address (possibly by doing what everyone else seems to?). Robbie
