And again, globally raise the timeouts.

follow start_applications/1 to app_utils

Merge default

Merged bug24907 into junk

Cope better with a server sending garbage during handshake

Merge heads

Added tag rabbitmq_v2_8_2 for changeset db6f2fb3b0ec

merge bug24917 into default

bug: code_change/3 should return {ok, State}

Merge bug24866 back into junk

Unmerge bug24866

Merged bug24866 into default

Merge bug24695

Allow specifying a default vhost

Auth mechanisms in URIs.

Junk bug24866

API change

Merge bug 24821

pmon is used by various of these modules...

merge bug23343 into junk

Use rabbit_net:ssl_opts on connecting, to 'expand' the option cacertsdir if present.

Rough sketch of exchange decorator concept.

Rename cluster config to runtime parameters.

handle rabbit_net:async_recv errors gracefully
plus some drive-by refactoring

Added tag rabbitmq_v2_8_1 for changeset ae6f225b4812

Some pluggable validation thing.

Added tag rabbitmq_v2_8_0 for changeset 9cbdf7e82983

Ignore this message.

Mgmt doesn't construct the name from peer_address and peer_port any more, we should generate a name here if there isn't one, full stop.

merge default into bug24152

increase timeouts

merge bug24605 into default (Move names for connections and channels from mgmt to broker)

Merge in bug24605 again.

Channel names.

Merge in default

Merged bug23114 into default

merge default into bug24152

Merge bug24668 (make consumer flow control more like credit_flow; improve performance along the way) (again)

Junk that, I merged to a head which got junked. Damn.

Merge bug24668 (make consumer flow control more like credit_flow; improve performance along the way)

Junking broken head of default

Whitespace change to make new head

Merging bug24152 to default

comply with modified notify_sent contract

merge default into bug23114

add rabbit_nodes to common

Merged bug24624 into default

Merged bug24670 into default

Merge bug24500

Merge bug24606

get rid of the redundant second argument to handle_wait_for_confirms

Merge bug24692.

refactor: inline

simplifying refactor

Move rabbit_misc:nodeparts/1 and rabbit_misc:makenode/1 into rabbit_nodes

Update copyright 2012

merge bug24336 into default

server_initiated_close() expects to be called from a cast not a call.

tiny refactor: 'case' -> multi-headed fun

Merged bug24664 into default

Provide a way to "server close" a direct connection.

Unbreak the network client. Oops. Tests pass now :)

deconvolute logic

The first problem pointed out in the previous commit no longer applies.

The only change in semantics is that wait_for_confirms_or_die will wait for
all acks and nacks before exit(nacks_received). It shouldn't matter much.

add timeout to wait_for_confirms[_or_die]

This also fixes two bugs:
1) Handling nacks never worked. It would do a gen_server:call inside a
gen_server:call and would just lockup. We circumvent the problem now, by
calling handle_close/4 instead of close/3;
2) In channels_manager, after the first erronous frame, the state would be
replaced by 'ok'. That's fixed now.

Merged bug24507 into default

hide specs for the two gen_consumer implementations

The docs for both of them still reference the behaviour which includes the
specs as part of its documentation.

Relax default timeout

Only invoke rabbit_channel:do_flow/3 when we are actually doing flow control...

WIP

I shall charitably assume that Matthias left that there to check we were paying attention.

keep up with server changes
rabbit_channel:do is no longer subject to flow control

merge default into bug24664

Merge bug24680

keep up with server changes

make tests work on R15

keep up with server API changes

Actually this is more reasonable.

Unbreak the direct client

Merged bug24659 into default

Portability: use temporary files instead of "sed -i".

The "-i" option isn't available on BSD systems.

Added tag rabbitmq_v2_7_1 for changeset 3e6a49651e8f

API change.

Propagate API changes.

Typo

Added tag rabbitmq_v2_7_0 for changeset bd1f29a08e42

cleanup after rpc client

And re-add broker stop after tests.

cleanup exchanges after tests

refactor

cosmetic

cleanup queues after use in tests

This fixes the original issue. I'll delete the exchanges as well, for
consistency.

better naming for randomly generated resources

Merging bug23326 to default

merge default into bug23326

Merge bug 24488 (Impossible to specify socket connection timeout with Erlang client)

expose socket_options in amqp_params_network

don't forget about timeouts with ssl

expose connection_timeout amqp_params_network option

Reduce multiheadedness

Merging bug24443 to default

Merged bug24499 into default

There should be no space there.

The set might be empty. Don't blow up.

merge bug 24499 into default (channel:wait_for_confirms is broken in the face of multiple=true)

Make wait_for_confirms not hang under the influnce of multiple=true.

Merged bug24453 into default

Don't include amqp_client.hrl; it's included transitively by amqp_gen_consumer_spec.hrl.

make tests a bit more comprehensive

add more checks to the tests

There was also a bug: "amqp://host" would parse to vhost "", instead of vhost
"/".

support mixed case schemes + more docs

support R12B-5 and better docs

Backed out changeset c4ac9194e945

rename url -> uri; do params_network conversion in amqp_uri

fix error reporting

remove socket_closing_timeout

refactor + handle entities + url tests

refactor

uri -> url

transplant the url parser from shovel

transplant uri_parser from rabbitmq_shovel

merge default into bug23187

remove channel and connection close timeouts

Added tag rabbitmq_v2_6_1 for changeset 493e002e47fa

Added tag rabbitmq_v2_6_0 for changeset d33ceb64fe87

Add support for credit reports to the erlang client. Just tells the consumer what it's been told itself, which seems par for the course here.

use confirms to ensure messages ended up in queues...
so we don't rely on 'purge' not overaking 'publish'.

use 'purge' instead of passive 'declare' to determine msg count...
since as of bug 24194 a 'declare' may overtake a 'publish'

Merging bug24230 to default

Merge in default

merge bug24132 into default

dependency on pg_local is gone

Move the pg_local management into rabbit_direct.

Export list/0 for poking around

Merging bug24344 to default

warning

support multiple consumers on the same process

Merging bug24018 to default

Rework return from new_connection so that we detect sooner if there's been a problem creating the connection

merge default into bug24018

Merge all the _exists events back to _created.

Oops, that got missed.

Merging bug23838 to default

start the infrastructure sup before the connection

I think the connection_sup processes are in the correct order for
automatic shutdown: channel_manager (with channel_sup_sup and all the
channels), gen_connection, connection_type_sup.

Channel_manager gets started by gen_connection:connect. We should
probably get rid of the chained call.

channels_manager (and channel_sup_sup) is started last in a connection

Don't start the channels manager before the infrastructure (and hence
don't thread its pid through 3 different processes). Just have
main_reader request the channel manager's pid on demand.

start channel_sup children in the right order

API changes, export the new behaviour.

Don't call rabbit_event:notify directly, we may not be a broker.

Use the new abstraction.

Introduce amqp_direct_connection:force_event_refresh(). It's a bit of a pity that we can't have this invoked from rabbit:force_event_refresh(), but the broker must not know it exists, and apart from that it's pretty clean.

Junk bug24132

Pre-junk this - pg2 is too slow

ccreated_event info item for the connection_exists event.

ipv6 + ssl test

add ipv6 test

support IPv6 address in Erlang client

Merge bug24284

silence tar

merge from default

Edoc doesn't break down anymore.

silence zips

silence unzip

add fullstop

update copyright year

merge default into bug24018

direct tests run again

(and covertly strip trailing whitespace)

merge from default

Direct tests still fail. There might also be a test or two missing.

As reported by Alvaro Videla, calls to monitor and demonitor should still be 'erlang:' prefixed for older versions of Erlang

Merging bug23117 to default

impossible case should cause an error

error, not exit

handle only one unassigned consumer at a time

merge from default

send basic.consume to consumer during rpc processing

fix a race

documentation improvements

less amusing direct_consumer behaviour

Merge bug24229 (again)

Merge heads of v2_5. Let's pick the correct tag for .hgtags, not that I think it makes any difference.

merge default into bug23187

forgot an unused From

merge default into bug23117

merge default into bug24229

remove From

merge bug24270 into junk (it's INVALID)

match the java client's behaviour wrt wait_for_confirms

writer should be intrinsic

Merge bug24229 to default

Reduce duplication

special close message when receiving a nack

refactor

calls and subscribes return ok | {ok, Result} | {error, Reason}

receiving a nack should also take down the channel

Just like in the other clients.

fix the channel killing itself on receiving a nack

app_initiated_close is represented just by its exit signal

update specs

don't double wrap with {error, _}

let's call it {error, _} instead of {shutdown, _}

cosmetic

Merging bug24138 to default

reorder clauses and complete specs

simplify

make timeout infinite