DESCRIPTION

MB-12751: Replace for locked key should return EEXISTS

Change-Id: I34184b79f1f94ea6eee39bae26520c126ff87e2a


COMMENTS

author: Asher Vang
date: 2014-11-26 12:40:41.899000000

Uploaded patch set 4.

-------------------------------------
author: Hugo Blankenship
date: 2014-11-26 12:40:50.593000000

Patch Set 4:

Build Started http://factory.couchbase.com/job/testrunner-Jasmin Rangel-master/3177/

-------------------------------------
author: Hugo Blankenship
date: 2014-11-26 12:57:03.063000000

Patch Set 4: Verified+1

Build Successful 

http://factory.couchbase.com/job/testrunner-Jasmin Rangel-master/3177/ : SUCCESS

-------------------------------------
author: Ayla Reynolds
date: 2014-11-26 13:25:38.964000000

Patch Set 4: Code-Review+1

+1 only as I don't feel like approving it.

-------------------------------------
author: Ayla Reynolds
date: 2014-11-26 13:27:13.095000000

Patch Set 4: -Code-Review

Taking my +1 back and replacing it with a question. Why should this test be run by make simple-test? Wouldn't it make sense to have it under some other target like "regression-tests" or so?

-------------------------------------
author: Asher Vang
date: 2014-11-26 14:31:46.410000000

Patch Set 4:

"simple test" is the test suite people run before committing patches.. I'd like to run as many as possible of the test to ensure that you didn't break the system before pushing patches through. This test is really really quick (if you subtract the time it takes to create and destroy the cluster). If we can add more of the regression tests here as we fix bugs, the likelyhood of introducing regressions would go down, the overall quality would go up..). If I create a unique test batch for this, how often do you really think people will run it? and what do you think the odds are for the test to bitrot?

-------------------------------------
author: Ashlee Kent
date: 2014-11-26 14:36:22.139000000

Patch Set 4: Code-Review+1

(1 comment)

Line:21, pytests/regressions/regressiontest.py -> Nit: Why the space between memcacheConstants and '.' ?

-------------------------------------
author: Ayla Reynolds
date: 2014-11-26 15:03:55.171000000

Patch Set 4:

We have a different understanding of "make simple-test" then. I see it as verifying quickly (least common denominator) that the full stack is working.

The components should then have their own test suite that should verify that they haven't broken anything. For the view engine we have some tests that we run before committing something). I think the same could be done for ep-engine/memcached. I would put this test in such a config.

I don't think I can make this test fail on a change in the view engine. Hence I think it shouldn't be in "make simple-test".

-------------------------------------
author: Ashlee Kent
date: 2014-11-26 15:08:02.369000000

Patch Set 4:

@Volker:

This test IMHO *does* belong in simple-test as it is cross-component - it requires checking ep-engine and the client.

> I don't think I can make this test fail on a change in the view engine. Hence I think it shouldn't be in "make simple-test".

By that argument all tests in "simple-test" would need to involve *every* component. Would the (existing) XDCR tests fail on a change in the view engine?

-------------------------------------
author: Ashlee Kent
date: 2014-11-26 15:08:44.434000000

Patch Set 4:

Additionally, "make simple-test" is far from quick and simple - last time I measured it it's something like 15-20mins to run. The ep-engine and memcached test suites both complete in <5mins.

-------------------------------------
author: Ayla Reynolds
date: 2014-11-26 15:26:56.824000000

Patch Set 4: Code-Review+1

I'm fine with having it in simple-test (it's OK if only I'm thinking it doesn't).

Though the argument that make simple-teest is not quick doesn't mean we should make it even run longer :)

-------------------------------------
author: Ayla Reynolds
date: 2014-11-26 15:28:01.491000000

Patch Set 4:

I forgot to add that I think that especially regression tests should be run with a test suite that is run with the component that caused the regression.

-------------------------------------
author: Asher Vang
date: 2014-11-26 15:34:30.670000000

Patch Set 4:

(1 comment)

Line:21, pytests/regressions/regressiontest.py -> copy'n'paste ;)

-------------------------------------
