[ 
https://issues.apache.org/jira/browse/HIVE-17982?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16317009#comment-16317009
 ] 

Alan Gates commented on HIVE-17982:
-----------------------------------

Peter, thanks for the review.

bq. adding more code between getting a free port and starting the metastore 
could cause racing condition problems
What race condition are you thinking about?  Each test class is run in a 
separate JVM, and each test in a class is run serially, so there's no 
possibility of a race condition between tests.  Something else in the system 
could be grabbing ports, but that seems unlikely.  We need a way to pick a port 
for both client and server.  Perhaps we could change startMetaStore to return 
the port it picked.  We'd have to check first that we always start the server 
first.  I was trying to keep my changes to a minimum so I didn't do this.  If 
we want to do that I suggest we do it in a separate patch. 

On the other comments, in places where the original was strange (like the 
try/catch blocks with assert(false) in the catch) I mostly left things alone.  
I agree it's goofy but it seemed better to leave things as they were rather 
than try to fix everything and possibly end up breaking something.  

On the original tables being bucketed, I don't think it matters.  That looked 
like copy paste code and I was too lazy to keep it everywhere when I ported it 
from SQL to the builders.  (I know this flies in the face of my comments in the 
previous paragraph.)



> Move metastore specific itests
> ------------------------------
>
>                 Key: HIVE-17982
>                 URL: https://issues.apache.org/jira/browse/HIVE-17982
>             Project: Hive
>          Issue Type: Sub-task
>          Components: Standalone Metastore
>            Reporter: Alan Gates
>            Assignee: Alan Gates
>              Labels: pull-request-available
>         Attachments: HIVE-17982.patch
>
>
> There are a number of tests in itests/hive-unit/.../metastore that are 
> metastore specific.  I suspect they were initially placed in itests only 
> because the metastore pulling in a few plugins from ql.
> Given that we need to be able to release the metastore separately, we need to 
> be able to test it completely as a standalone entity.  So I propose to move a 
> number of the itests over into standalone-metastore.  I will only move tests 
> that are isolated to the metastore.  Anything that tests wider functionality 
> I plan to leave in itests.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to