Eliminate javadoc warnings in com.rabbitmq.utility

Modify codegen to produce output with no javadoc warnings.

Merge in default

Eliminate JavaDoc warnings from ClientVersion generated java source.

Put arguments the right way round on assertEquals.

Split close processing (channel and connection);
- start and finish parts
- defer finish until processing response from the broker for application initiated close.

Merge bug24527 in to get clean compile.

merge bug24527 into default

Oops

Java 1.5 fix.

Revert properly.

Revert the removal of checkShutdown() in QueueingConsumer.

Backout changeset bbfc45f00750cfa5040af26df5d414a415c4ab19

Merge in default

Move channel shutdown processing to after close-ok reply received.
Move WorkPoolTests to com.rabbitmq.client.impl package space.

Merge default

There are different CAs in use now.

merge bug24527 into default

Make MulticastParams create producers and consumers, and thus prevent knowledge of its entire state leaking into MulticastSet.

Read performance test specifications from a JSON file. We lose the ability to run multiple brokers, but that wasn't very well done anyway. We can come back to it.

Allow strings to be quoted with "'" as well as '"'.

Deal with // comments.

Merged bug24910 into default

Add message to channel.close().

Merge in default

Use Connection.abort(…) instead of AMQConnection.close(…) in handleChannelKiller()

This actually makes the code simpler.

Remove JavaDoc warning, and put "DefaultExceptionHandler:" on logged message.

Rework handleChannelKiller to close channel only (if possible).

Merging bug24907 into junk

Added tag rabbitmq_v2_8_2 for changeset 9ee49999b983

Ignore garbage from server on first connect

Remove GRATUITOUS_WAIT and add more aggressive Requeue/Consumer tests.

Modify WorkPool so as not to complain so loudly if the channel is not in the pool.
Modify WorkPoolTests to check this correctly.
Small JavaDoc updates in the area.

Properly remove checkShutdown() and fix some JavaDoc bugs.

Merge in default

merge bug24881 into default

Increased TTL interval.

On my machine 1 sec interval is fine, but it start failing around half a
second. Hopefully this will be enough for the CI server.

queue leases with if_empty & if_unused option

Merged bug24876 into default

Support MultiAckEvery in MCM

Merge bug24725

test confirm for messages expired with x-message-ttl=0

test dead-lettering for x-message-ttl=0

make sure test doesn't get stuck on failure

test x-message-ttl=0 behaviour

merge default into bug24725

Minor tweaks to tests.

Rename package

Merge default in.

Eliminate warning from test

Finished simple call unit-test.

Update StubChannel

Merge default in

merge bug24845 into default

expand confirm tests to cover non-persistent m&i returns

refactor confirm tests and test more combinations of p&m&i

permit 0 in x-message-ttl, and test for that

Merge bug23463

Merged bug24835 into default

Correct call to handleAckNack on receiving a NACK.

Merge 24795 into default.

Eliminate a few unnecessary UnsupportedEncodingException references.

Merge in default

Eliminate import ..*; declarations.

Eliminate wildcard import in mainline code.

Fix noLocal parameter JavaDoc

String RPC uses UTF-8 consistently

Merge bug24827

Disk free space monitoring tests

a test

simplify further, and prevent NPE in String constructor on failure

reduce the execution time from 7.4 to 1.6 seconds

simplifying refactor

Added tag rabbitmq_v2_8_1 for changeset 8c29085e2aef

Added tag rabbitmq_v2_8_0 for changeset 2252ed45354f

Obtain rates in bytes/sec as well as msg/sec (not counting AMQP overhead)

Merge bug20337

remove useless test

Eliminate whooshing sound.

Stats should be instantiated immediately before you want to use them.

merge default into bug20337

expand deleted dlx test

add test for dead-lettering to a deleted exchange

Also, some of the other tests weren't binding queues when they should have.

refactor and thighten a test

get rid of PropertyFactory

It was unnecessarily complicated.

abstract nack and reject DLX test

Merge bug24756 into junk

More tweaks. This really needs to be a config file or something...

Add test for queueDeclarePassive(q).
-- QueueLifeCycle gets the test
-- mod to checkShutdownSignal to make it non-invasive on channel/connection state

Fixed grammar and some category errors in Javadoc.

Fix mistake.

Typo.

We, ah, already test this case elsewhere.

Merge bug24651

Attempt to document this outside of the impl package, I don't think people should be expected to look in there.

Add Javadoc

Add missing file from bug24747

unhang a test

This only became a problem when running all the java tests.

fix and refactor DLX tests

Some of the tests still used queue.purge to cause dead-lettering; we know use
queue ttls.

Merged bug24747 into default

add test for deleting a queue with a DLX twice

Merge in default

Adjusted test to limit hang if test fails.

don't dead-letter on delete and purge; update tests

Added test of logic

Make handleSocketTimeout() respond differently before negotiation completes

Merge bug24086

Merged default into junk

Unmerge bug20337

Merged bug20337 into default

1000 consumers require fiddling with ulimit, let's not...

Run the first test for a longer time, so in 2.7.1 we can see memory alarsm going off.

Fix FTBFS treating checked exceptions with the contempt they deserve.

One more scenario, tweak some params.

add test for empty-queue purge with dlx

re-add dead-letter-to-self test checking for the new condition

Also, I think purging an empty queue hangs.

merge default into bug20337

Start to push scenarios towards what we actually want to measure. Also MulticastParams should clearly not default to 10s.

More robustness in stopping the broker, better diagnostics if it doesn't start.

remove workaround now that the server is better behaved

That's too many scenarios and plotting them all on one chart is unreadable. Split out the before-paging and after-paging cases, that's all we care about.

Fix a couple of bugs introduced in stats. That reset idea is dodgy.

Don't return NaN.

MOAR DATA.

Scenarios with multiple phases; use this to test queue fill / drain.

Oops

Further naming

Different message counts for producer and consumer.

Now that we've got the name right, let's get the logic right too.

This is wrong apparently

Rename to message count.

Introduce a message limit flag, and make it so that if there are no consumers and a named queue we declare the queue anyway. No more qos1/tx2 trick!

Merge bug24234

Rearrange the JSON to be easier to work with.

More interesting numbers.

Underscores -> hyphens.

Make JSON easier to read during development.

Less vertical space

JSON

Rename a few things.

merge bug23710 into default

Revert a couple of dumb ideas.

A way to test different combinations of options without a cartesian product, and some tidying.

A bit more abstraction. And make the stuff we write to stdout be a bit nicer as a progress gauge.

Manage the broker lifecycle, and thus comparatively test various broker configurations.

merge default into bug20337

Multidimensional VaryingScenarios

Rate vs latency special case.

Split ProducerConsumerParams out of ProducerConsumerSet, introduce a couple of scenario types.

Move some totals up into the abstract Stats class, simplifying ScenarioStats. Thus move the totals stats out of Producer and Consumer and into PrintlnStats. This changes the behaviour of MulticastMain, but probably in a good way: before it would report total stats per producer / consumer, while amalgamating interval stats. Now it amalgamates both.

Get rid of mindlessly-generated getters

Start to refactor MulticastMain:

* Pull Producer / Consumer / Stats to top level.
* Make Stats abstract. Give MM an implementation that reports to stdout.
* Pull the Producer / Consumer start / wait / stop code into its own class.
* Introduce PerformanceMain, the (very) start of a performance testing thing.

Merge bug24723

Merge bug24606

dead-lettering to self no longer works

Another old copyright header

Update copyright 2012

Tidy test.

Write correct arraySize for real arrays.

Added test -- which fails at the moment :-(

Merge in default

Colocation

Rename and inline.

Stop channel consumer work pool after channel.close()

Junk bug24703

Begin test of call().

Complete testCallAfterClose().

Augment StubChannel to allow testOpen() to pass;

Rename test class

Abstracted the nacks/rejects test, removed the third basicGet() in the first test.

Merge bug24718

Chang testRollback{Nacks,Reject} to requeue before rollback.

Failing (in the case of txs) test case.

check that dlx is set if dl-rk is set

Also, one of the other tests wouldn't have failed when it should have.

Also fixed some minor cosmetic whitespace issues.

Merged bug24545 into default

Merged bug24701 into junk

Merged bug24648 into default

Merge bug24648 in

Update Test template for Maven bundle

Merge default in

This got renamed

Merge bug24648 in.

Adjust Maven pom.xml to remove commons-io and commons-cli from non-test scope.

Add mixed nack/ack + transactions test.

Add tests for transactional nack/reject.

Bringing back the correct default changeset

Mistakenly pushed changes to default

Added some tests for nack/reject.

check dlx routing key arg

merge default into bug20337

merge bug 24538 into default (WaitForConfirms with timeouts)

Merge bug24648 in.

Update org.commons.apache.* version ranges in maven bundle imports

Remove build dependency on org.apache.commons.io.* from maven bundle

Removed dependency on org.apache.commons.io.* from run-time code.

Close channel at timeout when waiting for confirms

Merge in default

Merged bug24687 into default

Merge in default

Eliminate RpcCaller abstraction layer; start unit tests of RpcClient.