The new sub command auth is not advertised in help
Running solr auth -enable without starting Solr throws a stack trace instead of printing usage.
Done, printed a message to the effect that Solr should've been started in cloud mode or zkHost should've been provided.
If you move the getZkHost(cli) check to after the credentials check, then it's a bit better.
it should be -Dbasicauth= without uppercase "A".
Method updateIncludeFileEnableAuth takes username, password as args but they are never used
Wrong code indent in SolrCLI#3629-3671
Guess we have to move the SOLR_HOME resolution lines higher up in the script
I found that using $SOLR_PID_DIR was much simpler than the $SOLR_HOME (which more or less point to the same location). On Windows, used $SOLR_TIP/bin.
Are you confident that this feature will have good enough quality to go in 6.6?
This is a new feature. So long as it doesn't trip up any existing parts of Solr, and it works for the cases we've tested manually, I am confident to have it in 6.6. Any bugs, if they escape attention, can be fixed later. Not putting it in 6x would delay the actual adoption by users, who are more likely, in the short term, to upgrade to 6.6 than 7.0.
I would expect it to be possible to cover most of the SolrCLI functionality in with unit tests.
Actually, I found it quite difficult to test the changes I've introduced to SolrCLI without writing some fundamental support to test external systems here. For example, I would've liked to test if the correct security.json is being uploaded to ZK or not. But without significant effort in building such scaffolding in our test framework, I couldn't see a way to test for that. Did I miss something obvious? Can you point me to any existing tests which I can derive any clues from? I didn't find the tests for the Examples very useful.
For this patch, I have tested manually on Linux, and still testing on Windows.