Download custom jars from s3 #315
Conversation
Both Solr and Yokozuna's custom class files must be built with JDK 1.6 so that it will work for JRE 1.6 and greater (official Solr 4.x builds against 1.6 so I'm following their lead). Instead of building these jars fresh each time, leaving them at the mercy of whatever javac happens to be on the official build machine, download them from s3. This way the jars can be built with the correct javac once, and then simply downloaded. Anytime any of the custom classes change the appropriate jar version must be bumped and the new jar along with its SHA1 should be uploaded to s3.
As long as Yokozuna is using Solr 4.x it needs to use JDK 1.6 for building the WAR just like the official Apache build.
|
Works for me: https://gist.github.com/bowrocker/9117066 Subsequent compiles without 'make clean' do not download them again. This is a good thing, but what happens if the jars change? Do we have to remember to do a clean to update? If so, perhaps a follow on is to check and track a checksum, and download only if it changes? |
|
Actually they should be downloaded again after clean because the files are not cached like the custom Solr package. They are small so hopefully that is fine for now. The jar are versioned and the only way they change is if the scripts are updated to a new version. To get the new jars you'll want to run a make clean followed by make. |
|
Makes sense to me, just checking. Things like that have bitten me in the past. I am 👍 |
|
Yea this is why I added the version. The yokozuna-1.jar will never change. If we add or modify handlers that will require the following steps:
|
This should address #314. See that PR or commit msgs for more info.
In order to test this you should run
make clean && rm -rf depsto make sure you get latest. You'll also want to wipe out your verify dir.