What we do is set up version independent symlinks for reference from elsewhere for installing the server jar or finding the client jar for JDBC ( https://github.com/apache/bigtop/blob/master/bigtop-packages/src/common/phoenix/install_phoenix.sh) . We also replace dependency jars in the Phoenix lib/ with symlinks ( https://github.com/apache/bigtop/blob/master/bigtop-packages/src/rpm/phoenix/SPECS/phoenix.spec) but this part can go away if the Phoenix lib/ goes away. An open question is what to do with the Phoenix binscripts. We don't do anything with them but I think launching sqlline etc. out of where Phoenix is installed will accidentally work. Keep these or remove them?
On Wed, Apr 22, 2015 at 9:39 AM, Andrew Purtell <apurt...@apache.org> wrote: > Not building a tgz will mess up Bigtop packaging, please don't rip it out > if it's not critical. > > > On Sat, Apr 18, 2015 at 2:20 PM, Nick Dimiduk <ndimi...@gmail.com> wrote: > >> Nothing filed; just asking for my own enlightenment. Let me see about >> putting our tgz's on a diet. >> >> Somewhat related: has anyone run make_rc.sh from a mac? It uses command >> syntax that apparently is not compatible with BSD tools. Should I point it >> at GNU tools, has anyone tried that? >> >> -n >> >> On Sat, Apr 18, 2015 at 12:36 PM, Jesse Yates <jesse.k.ya...@gmail.com> >> wrote: >> >> > Sounds like we can rip it out then - did you already file a jira? >> > >> > On Fri, Apr 17, 2015, 3:06 PM Nick Dimiduk <ndimi...@gmail.com> wrote: >> > >> > > My understanding is that we already run via standalone uberjar for >> > > sqlline.py and psql.py ... which may be a bug for folks who have >> deployed >> > > on versions of HBase/Hadoop we're not packaging against. The only >> things >> > in >> > > their class path is the HBASE_CONF_DIR and the client assembly jar. >> > > >> > > On Fri, Apr 17, 2015 at 2:14 PM, Jesse Yates <jya...@apache.org> >> wrote: >> > > >> > > > I think it was for convenience of packaging. While we have >> transitive >> > > > dependencies, they resolve when including the HBase/Hadoop >> classpaths >> > as >> > > > well. I think mostly this was to make it easy to run from the >> tarball >> > > using >> > > > sqline or the various python scripts in the bin/ directory. >> > > > >> > > > If no one is using them, then by all means, we should remove >> building >> > > them. >> > > > Or build them completely so we can run standalone (and then maybe >> > people >> > > > will use it). >> > > > >> > > > On Fri, Apr 17, 2015 at 2:06 PM Mujtaba Chohan <mujt...@apache.org> >> > > wrote: >> > > > >> > > > > Jesse I think you initially worked on Phoenix assembly tar and jar >> > > > > packaging. Any thoughts? >> > > > > >> > > > > On Fri, Apr 17, 2015 at 12:39 PM, Nick Dimiduk < >> ndimi...@gmail.com> >> > > > wrote: >> > > > > >> > > > >> Quick question: >> > > > >> >> > > > >> Why do we package up our dependencies in the TGZ? There's no >> Phoenix >> > > > >> executable, just a fat jar for the RS and a fat client jar for >> > > > >> applications. Why bother with lib and all this business? >> > > > >> >> > > > >> If we are trying to package up all our dependencies in the tgz, >> our >> > > > >> current >> > > > >> means are inadequate. A little shell gymnastics*** with mvn >> > > > >> dependency:list >> > > > >> shows me 201 total transitive dependencies for Phoenix, of which >> we >> > > only >> > > > >> package 38. >> > > > >> >> > > > >> So why bother? >> > > > >> >> > > > >> Thanks, >> > > > >> Nick >> > > > >> >> > > > >> *** >> > > > >> $ mvn dependency:list | egrep '\[INFO\] \w+' | grep compile | >> cut >> > > -d: >> > > > >> -f1-3 | sort | uniq | wc -l >> > > > >> 201 >> > > > >> >> > > > > >> > > > > >> > > > >> > > >> > >> > > > > -- > Best regards, > > - Andy > > Problems worthy of attack prove their worth by hitting back. - Piet Hein > (via Tom White) > -- Best regards, - Andy Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)
