Improve UnknownObject observability

Add an UnknownObject.toString() method, so that UnknownObjects expose
their state for debugging purposes.

JIRA: BGPCEP-981
Change-Id: I414e5cbd6967811bb6174f6d289f88fba1d1db08
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Do not reference opendaylight.org/xmlns/blueprint/v1.0.0

bgp-rib-impl is not using any Blueprint extensions, hence there is no
need to bind to this namespace.

JIRA: BGPCEP-958
Change-Id: If6c1a3ddaad47f6f8e0df4d0b8accbe21d3b625c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Switch bgp-topology-provider to OSGi DS

The container we use here is extremely simple, migrate it to Declarative
Services.

JIRA: BGPCEP-959
Change-Id: Ib7f6f7f1925949f2b94ac8a4427c36dac838fa59
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 0.16.7-SNAPSHOT

This starts the next development iteration.

Change-Id: I615061b5db91275b19a617e858c4fd04d2218631
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Intern constant QNames

We really want to reuse global constants here, make sure that happens.

Change-Id: I7cc78f29fbfae873b3920b3684bde8c331a3d0af
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Update withdraw()/advertize()

We have a simple boolean dispatch here, mark the boolean and use a
a simple return expression.

Change-Id: I5f45d160ef30d9b307a2e28578609dbeca27bd95
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Use DistinctNodeContainer.getChildByArg()

The check implied by Optional.get() differs only by the exception thrown
-- and VerifyException is more appropriate. By same token we can
side-step the Optional and use childByArg() instead.

Change-Id: I317577ef635c621be360297a2b5b376edbbbd2d4
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Cleanup AdjRibOutListener a bit

We have a few single-caller methods which can be inline a tad
more efficiently.

Change-Id: I79d85e30a1deb05e0341b8b20ee6925182287dbf
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Prefix/PathId leaf names are constant

With the switch to DOM processing, these have become proper constants,
make sure we use their shared versions and make make a few methods
static as a result ot that.

Change-Id: I09115b77ea33bbd9f0427f923e5bfa7b21c1127d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Revert "Flush session on AdjRibOutListener convergence"

This reverts commit b2c79a62bd1d45a7b9ff83332386c30d3ea1b8d3, as it
seems to be causing problems in CSIT.

Change-Id: Id023247a624c4e0459e0573a9105986a96cb3351
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix PCEP registry warning

We are logging getClass(), which is a bit confusing, as that is not the
key we are looking up. Make things obvious by logging
implementedInterface().

Change-Id: I80540b21554de1909a8e77930306b77ced3c9be3
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 0.16.6-SNAPSHOT

This starts the next development iteration.

Change-Id: If1238818b8e7cbd891c112f77d42bcdf4daa1fe8
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix test tool shading

We are shading a number of jars which provide services, which are
overlapping. Make sure we use ServicesResourceTransformer to merge these
entries.

Change-Id: Id5fab04d52575dc2e68526fe9c0b42dee6addef7
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Cleanup pcc-mock

Eliminate duplicate version specification. Also cleanup Mockito usage
and builder layout.

Change-Id: I579d39f1a101884be5e83a3ebd247d5a9cb951ec
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Migrate bmp-impl tests

Do not use MockitoAnnotations.initmocks(), but rather
MockitoJUnitRunner, which has further benefits.

Change-Id: I85589b5264f62a07bbe87f13daf899fa3e947aed
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix BmpDeployerImpl activation

Ie114a01b8ec1f87926f752b31cd7762a542e5db3 broke activation by failing to
annotate BmpDeployerImpl. This patch adds the appropriate annotations.

JIRA: BGPCEP-953
Change-Id: I8a59fc6977ddcd3e4861d02ff8df7f61820aced9
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 0.16.5-SNAPSHOT

This starts the next development iteration.

Change-Id: I81e654de999396cdb688307acdd4de25b0ee640d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix PcRequest Message Parser in presence of LSP

Since RFC8231, PcRequest messages can include a LSP Object.
However, the insertObject() method of the PCEPRequestMessageParser class
does not handle LSP Objects.
Thus, when a PCC sends a PcRequest message with a LSP, ODL throws a
Java Exception about Unknown Object and sends a PcError with RP
Missing Object type.

- override initial PcRequest parser with a new StatefulPcRequest
  message parser in the ietf-stateful package
- add support for LSP Objects when validating the PcRequest message

JIRA: BGPCEP-975
Signed-off-by: Olivier Dugeon <olivier.dugeon@orange.com>
Change-Id: I525cd933237bf3633da8c2ff854989b1f9968aa8

Bump upstreams

Adopt netconf-2.0.5 et al.

Change-Id: I54c1d4f581a4701def076b590c6215d181edecd0
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix parsing issue of PcReport Objects

StatefulPCReportMessageParser() class assumes that PCEP Objects in the
PcReport message are sent in a certain order. However, the order of
Object in the PcReport has changed between old draft version and final
RFC8231.

Indeed, as per RFC8231, the PcReport is composed of:
 - [SRP], <LSP>, <path> where:
   - <path> = <intended-path>
              [<actual-attribute-list><actual-path>]
              <intended-attribute-list>
   - <intended-path> = ERO
   - <actual-attribute-list> = BANDWIDTH, METRICS & <actual-path> = RRO
   - <intended-attribute-list> = LSPA, BANDWIDTH, METRICS, IRO

While in old draft version, <intended-attribute-list> was placed just
right after the <intended-path>.

Thus, the state machine should be flexible enough to accommodate to
PCCs that continue to use old draft and PCCs that are compliant to the
RFC8231.

JIRA: BGPCEP-974

Signed-off-by: Olivier Dugeon <olivier.dugeon@orange.com>
Change-Id: I97c4e339d8fe66d5e6a10fbb7055c91941351eda

Bump versions to 0.16.4-SNAPSHOT

This starts the next development iteration.

Change-Id: I1c075a44da98ea9f0930c3736ce25de207a8d718
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump upstream versions

Adopt netconf-2.0.3 et al. Unfortunately maven-dependency-plugin's logic
seems to have shifted, hence we disable that for now.

Change-Id: Id2e47947f544aab08ba32759199a907d87616a76
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 0.16.3-SNAPSHOT

This starts the next development iteration.

Change-Id: Ie0b44359c2dbb9ac2cf0f215d9b0f0532f44a576
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump upstream versions

adopt yangtools-7.0.4, mdsal-8.0.1, controller-4.0.1 and netconf-2.0.2.

Change-Id: I9c4264b4bf81a596365d65f412d3f6defec06ea8
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Improve Pcep Server

Some PCCs send a PcRequest message with Metrics and/or Bandwidth set
with value equal to 0 which result to a NO_PATH when calling path
computation. Indeed, it is impossible to determine a path with a metric
less or equal to 0.

This patch set skip Metrics and/or Bandwidth equal to 0 when building
the list ofi constraints for the path computation algorithm.

Change-Id: I3b343e67fc35f3393d4ed136aa4224082823f066
Signed-off-by: Olivier Dugeon <olivier.dugeon@orange.com>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Redefine AbstractMessageParser object parsing

All AbstractMessageParser.validate() interact with the list of objects
by peeking at the first element and removing it when handled. While
the List contract supports it, there is no efficient implementation:
LinkedList is wasteful with memory (and its locality), ArrayList ends
up shifting arrays all the time.

Deal with this dilemma by recognizing we are dealing with a Queue, and
realize it through ArrayDeque. Let users deal with the logic rewrite,
which ends up being rather sensible: they are always looking only at the
first element anyway.

This also exposes a few places where we can run into runtime exceptions
-- mark those with FIXMEs for future improvement.

JIRA: BGPCEP-976
Change-Id: I882c7208dddd61f368499fd661e6d090752d144d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Cleanup ReoptimizationBandwidth reference

We have this class imported, use short name to reference it.

Change-Id: Ie6aee18db947f7d806b43b69df8548323658ce26
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Modernize AbstractMessageParser

- use List.of(item) instead of Collections.singletonList()
- reuse requireNonNull()'s return
- reformat builder usage
- use Optional.ifPresent() to propagate present values

Change-Id: I20022f19c80f2ca832f47914ac801fca124f568e
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Do not wipe data in ApplicationPeer

Unfortunately the wipe here is causing lifecycle issues, as there is a
pre-populated structure we rely on when things start coming in.

Change-Id: Ib58bf34e96e68e0db2fd3a8e580cade0d01099c0
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Optimize uint interactions in bgp-linkstate

Use writeUint{8,16,32} instead of manual conversions. Also protect
a hexDump() call. Also add obvious ByteBuf sizing hints.

Change-Id: I462b9d3d566efc2bee020df15ac08ed6c89dea8d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Eliimnate use to Uint8.toJava() in rsvp-impl

We have ByteBufUtils.writeUint8(), which is a tad simpler, use that
instead of dual conversion.

Change-Id: Ibd58934382e9244a0428c68722ef59611801f94b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix LinkstateGraphBuilder.ipv6toKey()

We are picking 8 most significant bytes of the address, which is
contrary to what the comment says. Add an explicit test and fixup
the implementation.

Change-Id: Ie3817bde8be858c0a5031b643994645aa61932df
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Do not copy Ipv{4,6}Addresses

Ipv{4,6}InterfaceIdentifier are specializations of Ipv{4,6}Address, hence
we do not need to copy them.

Change-Id: I23ba5837cf315e38a1c0dd57f51a88dee35d9e91
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove unneeded wrapping

We are re-wrapping an IpPrefix, let's just reuse it, improving clarity
and performance.

Change-Id: Ib27930c3ab998647a969f45afceea8ff74c93519
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove a blind cast

We have a null check here, let's use an instanceof check instead,
which is essentially the same thing, except more type-safe.

Change-Id: I2538e4e2f58513aa51bcefe26d4e55ad8ab59283
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Improve Ipv{4,6}Address handling

Ipv4Address can actually contain a zone, which would break prefix
creation. Use IetfInetUtil to convert address to prefix.

Change-Id: I5434d8c9b3164f6cee8c47b44a65bad5e02abf8f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Simplify LinkstateGraphBuilder.getLinkAttributes()

If have too many if/else statements here and a blind cast. Refactor
checks and use a simple return.

Change-Id: I158914f2e972f5f0265438f51eb287076b436459
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Improve withdrawn routes parsing

We are using readBytes() and release() to synchronously process a
buffer. Use readSlice(), which side-steps refcount manipulation,
making things a tad faster.

Change-Id: I89886868a2f45988620c6b8be8fb5a85c9df4ced
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix potential ByteBuf leak in bgp-inet

We are using ByteBuf.readBytes() to frame a parsing chunk, but
we do not call ByteBuf.release() on it. Use ByteBuf.readSlice()
instead.

JIRA: BGPCEP-973
Change-Id: Ib0e6be0d8bca3518f0f318f137ca1a05196f84d6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix potential ByteBuf leaks in bgp-l3vpn

We are using ByteBuf.readBytes() to skip over some content and also
to frame it for parsing. Use ByteBuf.skipBytes() in the case of former
and ByteBuf.readSlice() in case of latter. This speeds things up and
plugs a potential refcount problem.

JIRA: BGPCEP-973
Change-Id: Ib9bf406a5e9de9b5e122c3fb3323a4b772a1a798
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix potential ByteBuf leak in bgp-inet

We are using ByteBuf.readBytes() to actually skip over some content,
use ByteBuf.skipBytes() instead.

JIRA: BGPCEP-973
Change-Id: Ia33e199d63ea05fe4d3383384970f7aaf4441757
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix potential ByteBuf leaks in bgp-labeled-unicast

We are using ByteBuf.readBytes() to actually skip over some content,
use ByteBuf.skipBytes() instead.

JIRA: BGPCEP-973
Change-Id: I1bcdcc47b7ac6352e5bdc80064d40cf4692e7eb6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix potential ByteBuf leaks in pcep-base-parser

We are using ByteBuf.readBytes() to actually skip over some content,
use ByteBuf.skipBytes() instead.

JIRA: BGPCEP-973
Change-Id: I0163acd3f0d59e6bcb272af639092f2e1b74ed5b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Use ByteBuf.readSlice() in bgp-mvpn

readBytes() returns a ByteBuf, which needs to be be freed eventually
through ByteBuf.release() -- which we are not doing. Use readSlice()
instead.

JIRA: BGPCEP-973
Change-Id: Ide5af583cb33823a941093e6f9e7d5b61b4d2f3c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 0.16.2-SNAPSHOT

This starts the next development iteration.

Change-Id: I92e816000f8eff23e072c0386311516ad9c25f36
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Activate Segment Routing by default

Since Aluminium release, Segment Routing was wrongly disabled by
default, following the commit that update SR to latest RFC.

In previous release it was activated by default and the documentation
let users thinks it was enabled by default. This commit revert the
previous behavior.

JIRA: BGPCEP-972
Change-Id: Ia8c36e43e026dc576a6ef7b0f2faeac1bc9bef65
Signed-off-by: Olivier Dugeon <olivier.dugeon@orange.com>

Handle SID Index in SRGB block

When building Graph from BGP-LS, Segment Routing Global Block (SRGB)
advertized with a SID Index instead of an MPLS Label thrown a Java
Exception in getVertex() function. This commit allows the support of
both SID Index and MPLS Label for SRGB lower bound value.

JIRA: BGPCEP-970
Change-Id: I14cf39831f5ef47f7e80e91659ae6daeeacb1b91
Signed-off-by: Olivier Dugeon <olivier.dugeon@orange.com>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Optimize getVertex() a bit

Using Uint32 for AS number allows us to side-step a few conversions,
making the code a bit more direct.

Also we have type-safe versions of IPv4/IPv6 routerIDs, do not
instantiate new addresses via Strings which is relatively slow due
to the need to check values.

Change-Id: I170255815bb05628fff2a1baa5df0d6f30c1f9dd
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Do not store ConnectedGraphProvider

We are only acquiring ConnectedGraph, eliminate unneeded field as
pointed out by Eclipse. Also use static Preconditions.checkArgument()
reference.

Change-Id: Id1534fe204e47311ed1b10ee715acfac1d638b5f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Optimize LinkstateGraphBuilder.getVertexId()

Eliminate the use of a temporary Long, reducing the number of type
conversions we perform.

Change-Id: Iff20437ae066ba4618945c7c0230e5fd042d1f03
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix address handling in LinkstateGraphBuilder

When building Graph from BGP-LS, IPv4 addresses are wrongly converted
to Edge Key i.e. addresses greater than 128.0.0.0 thrown a Java
Exception.

This commit change the way Edge keys are computed from both IPv4 and
IPv6 addresses, eliminating Inet6Address.getByName() and using
efficient conversion utilities.

JIRA: BGPCEP-971
Change-Id: I3d00a92713675974dcce33e864dd23e93b0086f2
Signed-off-by: Olivier Dugeon <olivier.dugeon@orange.com>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump netconf to 2.0.1

Pick up latest upstream fixes.

Change-Id: Iefa625d68d5639531046e672633e05a78f00e495
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 0.16.1-SNAPSHOT

This starts the next development iteration.

Change-Id: I229deab151d5cd91f7d6dd5414886375f7a0f117
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Use released netconf-2.0.0 artifacts

Do not use snapshots, as we have released artifacts available.

Change-Id: I9aca9f4f8abe8b6e37bc9196821837d09e381c4f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Use controller-4.0.0 artifacts

Do not use snapshots, but rather released versions. This brings us
closer back to proper integration on releases.

Change-Id: I51824e3090966cc8836bc86ba09f1f1053139a97
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove blueprint from bgp-openconfig-state

We have some amount of configuration here, but that is rectified really
easily. Move the configuration for Config Admin and use OSGi DS instead
of blueprint.

The test suite is updated with a few FIXMEs around predictable
execution.

JIRA: BGPCEP-956
Change-Id: I3b0fe17a3dc1b685698730f897ae77c470a87ac8
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Clenaup use of MockitoAnnotations

Use MockitoJUnitRunner.StrictStubs instead of
MockitoAnnotations.initMocks(). Also perform a few drive-by cleanups as
well.

Change-Id: Iebee31479c88c8e492b92591f6047b5e8442035b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove ServiceLoaderBGPExtensionProviderContext

This class is used only in tests and it wreaks havot with its hidden
global state. Remove it and replace its users with localized
equivalents.

Change-Id: I59f2747c0ad65ff334c5b7ea2b2af6308fdc503d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Activate extension in bgp-testtool

Our extensions are nicely decomposed and can be activated through
ServiceLoader, do that in testtool.

Change-Id: Ie794de019cac527e8b546f67ed05562744c07680
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Do not override compiler configuration

odlparent's compiler configuration is in many ways better than what
we are doing here, just remove the configuration.

Change-Id: If95cf81dda6f65723effba833bb6fedee5751d12
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Inherit dependency analysis

Our dependency analysis has been upstreamed to odlparent, remove
unnecessary plugin execution (which is now a duplicate).

Change-Id: Ie3abbb27dfff6bff9374dfdb0c695b66bf53b086
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump argparse4j to 0.9.0

https://github.com/argparse4j/argparse4j/releases/tag/argparse4j-0.8.0
https://github.com/argparse4j/argparse4j/releases/tag/argparse4j-0.8.1
https://github.com/argparse4j/argparse4j/releases/tag/argparse4j-0.9.0

Change-Id: I8c817955cd71b5bb1c9b363deca7ce36c27c17da
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Use proper extension loading in bgp-rib-impl tests

Do not use global state in ServiceLoaderBGPExtensionProviderContext(),
but rather a normal ServiceLoader operation.

Change-Id: I36975d09cda8ff8e8e6815b15811b0aa56ade8f6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Eliminate use of global service loader state

Use normal service loading and clean tests up a bit.

Change-Id: I9c5e09ee8e9cb0d0158ca8c1a82e2e244c7a6e60
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Refactor (Ll)GracefulCapabilityHandlerTest

Clean up utilities that we are using and stop using global state.

Change-Id: I6f8ff0095ba0b5e1bdace775f909c91764721b62
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Refactor OpenTest

Use ServiceLoader to find extensions and take advantage of newer Java
utilities.

Change-Id: I6ed1c19ffbf85c484e0326942650256e80335222
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Do not use ServiceLoaderBGPExtensionProviderContext

Update extensions' tests to not use global state, but rather activate
themselves.

Change-Id: I0ce67fbd9ccba6c27c6217193a33c4c013ab9bdc
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Refactor BGPParserTest

Do not use global state and improve structure a bit.

Change-Id: I3416d5ca93d754d29f0e47abb733e0f1b6f94e9c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Refactor NextHopAttributeParserTest

Use ServiceLoader and assertThrows() to make the suite more reliable.

Change-Id: I02392e48767933e43e6d2efaf756fb7d5267db4c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Refactor OriginAttributeParserTest

Do not use global service loader state and use proper assertThrows().

Change-Id: Ibc304b985ab1d413d820402e5aff29f29979ead6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Reduce use of ServiceLoaderBGPExtensionProviderContext

Use proper serviceloader to load up the resulting consumer context,
rather than relying on magic global state.

Change-Id: Ic541ac82297558b745c5ded0ef156c2149586858
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Refactor AbstractMessageRegistryTest

We have a few weird asserts and global state going on here, fix that up.

Change-Id: Ib795fa7893eddb95bf0bb277bf216fec9d800cd6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Refactor SimpleRegistryTest

Do not use ServiceLoaderBGPExtensionProviderContext due to state
proliferation.

Change-Id: If0ddb86ee3aa76f95adc7aa00b06fa81b0f57dd0
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Refactor PathAttributeParserTest

We have a few interestng asserts here and we are using global state,
factor that out.

Change-Id: I93fd890baacb7fb8a22acd244341f6491c4efe43
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Refactor ParserTest

Do not use ServiceLoaderBGPExtensionProviderContext() and use
assertThrows().

Change-Id: Idba5dd5e58deea555dfa7673c88fc385c77889ed
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove ServiceLoaderPCEPExtensionProviderContext

We have DefaultPCEPExtensionConsumerContext which serves the activation
purposes without having global state lurking in singletons. Remove
ServiceLoaderPCEPExtensionProviderContext and replace its use in tests.

Change-Id: Ib09c7f93ad00b4cfb2cc293c0fa65004749b1fab
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove ServiceLoaderRSVPExtensionProviderContext

Carrying global static state is ill-advised, remove
ServiceLoaderRSVPExtensionProviderContext as it is superseded
through normal injection mechanisms.

Change-Id: I0996b3fb5dc69153def89fd4e57beea4c9e4054b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Split out PCEPExtensionProviderActivator

This is a really simple extension, split it out into a component.

JIRA: BGPCEP-968
Change-Id: I9d4abb0c54f169ef6715db8c1243cc008271b710
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Split out PCEPExtensionProviderActivators

PCEPExtensionProviderActivator is a simple whiteboard pattern, split
the activators into OSGi DS components.

JIRA: BGPCEP-969
Change-Id: I24feced4a4af71ab35b831a8b99f8dd1125cea7f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Migrate pcep-spi to OSGi DS

This is a simple whiteboard, refactor it to use OSGi DS (and others).
The changes here cascade through all extensions, making them properly
stateless.

JIRA: BGPCEP-961
Change-Id: I1e8f253f2ea7a2eef494c84612a867b6318a93c6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove a superfluous service specification

DefaultRSVPExtensionConsumerContext directly implements the service,
there is no point in spelling it out.

Change-Id: Ie97a021375d1db9017416764e7d337babc5b5800
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Split out BmpDeployerImpl

We have a rather contrived service routing here, ditch it in favor of
OSGi DS. Also eliminate BmpDeployerDependencies, as they are just
getting in the way.

JIRA: BGPCEP-953
Change-Id: Ie114a01b8ec1f87926f752b31cd7762a542e5db3
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Split out BmpDispatcherImpl

The dispatcher is a simple component, split it out into OSGi DS.

JIRA: BGPCEP-953
Change-Id: Idbdfbfdc1730bfbbdc44ed903584fb307dd4aeb6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Split out DefaultBmpSessionFactory wiring

DefaultBmpSessionFactory is a simple service, use usual DI frameworks
to activate it.

JIRA: BGPCEP-953
Change-Id: Ie4c120b652f8e60e0f451052201ec355ab4f61be
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Simplify peer group check

We do not need to perform an explicit nullness check, as we can simply
call equals on a constant.

Change-Id: Ibb713ea423dcc7b17e4a9a90541cd475bf03de35
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Split out BGPDispatcherImpl

BGPDispatcherImpl is a rather trivial component, split it out of
blueprint to make things a tad more clear.

JIRA: BGPCEP-958
Change-Id: I8d01ced1ce17ac5d52d78171d17bba39918c01c8
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Split out OSGiBGPStateConsumer

BGPStateCollectorImpl is a weird bit of code. We essentially have an
OSGi whiteboard pattern going on here, except we drive it through an
otherwise-unused BGPStateProducer interface.

Turn GPStateCollectorImpl into a @Singleton for things like Guice et al,
but it will need some more integration work to get it operation.

Add OSGiBGPStateConsumer which is a trivial OSGi DS whiteboard consumer
and thus eliminate a chunk of blueprint wiring.

JIRA: BGPCEP-958
Change-Id: I92fd2ecb0ef3f1ecb1069536c8b2a4d6261f23c4
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Split out StrictBGPPeerRegistry

StrictBGPPeerRegistry is a very simple service, split out of blueprint.

JIRA: BGPCEP-958
Change-Id: I51ac52bb979e727d510a6b56f5e56d020b01bcff
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Update initial convergence message

This is a slight improvement on messages logged after commit on
initial convergence.

Change-Id: Ic70949d69d4683fe4c9e0e4018e86f88af244106
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Clean EffectiveRibIn on ApplicationPeer convergence

If we converge with empty application peer RIB we should make sure the
operation effective-rib-in is wiped.

Change-Id: I6c7536406888e1d04f3624a8b3d2450746bd4cd8
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Flush session on AdjRibOutListener convergence

AdjRibOutListener should send out a flush session when it reaches
initial convergence without any data.

Change-Id: I3844a2912e0d23f679c4fcb620c8da451e6958c1
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Convert pcep-server-provider to OSGi DS

The blueprint here is a very simple single component, refactor it to use
OSGi Declarative Services instead.

JIRA: BGPCEP-963
Change-Id: Id396abaf268fce5ec01cea0e96a1d834551cb852
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Convert bmp-spi to OSGi DS

Follow the same pattern as we've done in bgp-parser-spi, ditching yet
another BluePrint container.

JIRA: BGPCEP-950
Change-Id: I9a787edf34fe9a6fb667b2f7f63d583f387a2a37
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Use mdsal-8.0.0 artifacts

mdsal-8.0.0, switch to using released artifacts.

Change-Id: I932810194ac4293ce584c4f317f1794c278b5f7d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add onInitialData() callbacks

DOMDataTreeChangeListener provides an empty-source callback for initial
convergence. Add instances of it along with FIXMEs to implement it.

Change-Id: I661622eecbd4a137bbac1e83d8209cc5a95a4379
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove unneeded references to ODL blueprint extension

We have a few blueprints which are not using ODL extensions, yet
they are declaring a dependency on it. Remove such references, making
future migration more obvious.

JIRA: BGPCEP-881
Change-Id: I4c5fffb4806f0beade7f7d9533b5ff6cef48f3d0
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Convert bmp-parser-impl to OSGi DS

BmpActivator is essentially a simple whiteboard dispatch, let's express
it that way, eliminating some blueprint wiring.

JIRA: BGPCEP-951
Change-Id: I77b4d0d3a53fb73adb72c8affb84e0c593919675
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Convert programming-impl to OSGi DS

This is a rather simple blueprint, convert it to OSGi DS.

JIRA: BGPCEP-922
Change-Id: I33b455490123cad6d20fb735ab72177b62d6cc09
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump MRI upstreams

Adopt:
- odlparent-9.0.2
- infrautils-2.0.2
- yangtools-7.0.3
- mdsal-8.0.0-SNAPSHOT
- aaa-0.14.0-SNAPSHOT
- netconf-2.0.0-SNAPSHOT
- controller-4.0.0-SNAPSHOT

Also adjust to API changes, mostly coming from yang-data-api, but there
is also a change in mapping for attributes-{reach,unreach} children,
which no longer are based on augmentation path.

Since augmentation sorting has been relaxed, this wrecks havoc in
odl-pcep-sync-optimizations.yang, which relies on the (illogical)
sorting rules of previous implementation. These are nailed down via
augment-identifier extension and sorted in a more logical way.

Change-Id: I54f7637a9443c9c577c6f397fe102a461f617229
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Convert algo-impl to OSGi DS

This is a very simple component wiring-wise, convert it to OSGi DS.

JIRA: BGPCEP-925
Change-Id: Idbad3b4fa4c92d41b6debe02865b59154c32f5bc
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>