git.opendaylight Code Review - odlparent.git/log

BUG-8923: create karaf-util

We need to use opendaylight-karaf-empty for SFT, but doing so
directly leads to a circular dependency, as karaf-plugin needs
sft (which needs karaf-empty, which needs karaf-plugin).

Resolve the dependency cycle by breaking out the single class
karaf-plugin requires into a separate artifact.

Change-Id: Id0649dd8203087ed82c1b936428cb6d9295f2957
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 5d3c00c2b98de5ebb4a7eda6e5cc0734c86b74f8)

BUG-8923: Bouncy Castle is provided

We need to exclude the Bouncy Castle artifacts from generated feature
descriptors, since they’re provided in our startup configuration.

Change-Id: I26814b09efa37d12c1e567e5622f509eea9241be
Signed-off-by: Stephen Kitt <skitt@redhat.com>
(cherry picked from commit 580bee99d255788f8e0f879075b8a12cb92c607f)

BUG-8923: package javassist in a feature

Javassist is an optional dependency to netty and is required by
binding mdsal. This can lead to re-wires when mdsal is pulled in
after anything using odl-netty-4 is installed.

Change-Id: Ia5135dfd27dd6227d9897d1a2ab24e0078673d8c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit e1b90dbbcdf55251192c2bf5b34c5a6f61e58590)

Fix tagGz property

karaf.archiveTagGz should be karaf.archiveTarGz, as used in the plugin
configuration.

Change-Id: I7173180a0e602867bd6a9c5b9fe1ac41a0a4ab6a
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump immutables to 2.5.6

General house-keeping to keep us uptodate.

Change-Id: Ic8ce978f8274701677997ac2987ec483816d7842
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add commons-text declaration

commons-lang3's WordUtils is deprecated in favor of commons-text
version. Downstreams should be able to migrate without having to
deal with versions.

Change-Id: I592f92b79372a21a8ada3225d05bed410a00363b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump javassist to 3.21.0GA

https://github.com/jboss-javassist/javassist/compare/rel_3_20_0_ga...rel_3_21_0_ga

Change-Id: I56af3a4a7ecacd919300e5c2a0332e749b1ed708
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump akka to 2.5.3

http://akka.io/blog/news/2017/04/13/akka-2.5.0-released
http://akka.io/blog/news/2017/05/02/akka-2.5.1-released
http://akka.io/blog/news/2017/05/24/akka-2.5.2-released
http://akka.io/blog/news/2017/06/19/akka-2.5.3-released

Also leaks sun.net from the JDK, as it is required now :(

Change-Id: I65ab6b1a2024084281774cfb17ea1bdd54cd1848
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump jackson to 2.8.9

Version 2.3.2 is really ancient, bump it to 2.8.9.

Change-Id: I92b5aa15c49fe0742f0069a88328e700d2e04f82
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add stax2-api declaration to match jackson

stax2-api is useful for others even without a provider, but
we cannot use 4.0.0 (as we'd like) because jackson has not upgraded
to it.

Change-Id: I69c15a360c2624fb842d4512accad67769de5d1d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add Javadoc links for third parties

Add some links for third party libraries declared as dependencies in
odlparent-lite so that those links are automatically added to the Maven
javadoc:aggregate or javadoc:javadoc goals. The list is not
comprehensive, it is just an starting point to continue adding more
links.

The projects/modules/aggregators referencing odlparent-lite will inherit
these links.

The <detectLinks> flag is set to false until [1] is resolved.

[1] https://issues.apache.org/jira/browse/MJAVADOC-427

Change-Id: I56a6eabe0ba7b9f32ffed7748ab13a79aa6bc2e1
Signed-off-by: David Suarez <david.suarez.fuentes@gmail.com>

Revert "Add an origin ID from the MDC into ..."

"... Karaf logs" until we agree on terminology etc.

This reverts commit 03609ff917ea5a187664614252a3e777108cb2ab.

Change-Id: I947fb2e44622f33e71614a66cf2ded816a61b26c
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Remove logback.xml

as well as the reference to it in etc/custom.properties

as discussed on https://lists.opendaylight.org/pipermail/odlparent-dev/2017-July/001262.html

Change-Id: I4423d3950e3d5f23be6d59423ef8b39abd0866a8
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Add an origin ID from the MDC into Karaf logs

see https://git.opendaylight.org/gerrit/#/q/topic:mdc related changes

and read https://logback.qos.ch/manual/mdc.html
or https://logging.apache.org/log4j/2.x/manual/thread-context.html

This change will show the OID in all logs, for example like this:

2017-07-27 23:14:16,034 | WARN  | 0000000000000 | ion(1)-127.0.0.1 |
InfrautilsIT                     | 93 -
PAXEXAM-PROBE-41ce6a29-abaf-44c0-8044-71d3e2bf99da - 0.0.0 | Ho, ho - do
you also see this funky Origin ID at the <== start of this log message:
0000000000000

If there is no originID available, then the new column will just be 13
empty spaces (for correct padding, like all other columns in the log),
like this:

2017-07-27 23:14:16,554 | ERROR |               | 0e6-c784b3cb4f5b |
core                             | 6 - org.apache.karaf.features.core -
4.0.9 | FrameworkEvent ERROR - org.apache.karaf.features.core

Change-Id: If18e1d35d71da5bd5c0fee5f2212915cccb0992e
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

BUG-7464: do not package upstream triemap

yangtools carries a reworked fork with a large number of bugfixes
and it is no longer using upstream. Remove odl-triemap, letting
downstreams consume the yangtools fork.

Change-Id: I5c04c5bed193594a57d4845f857ab62637b70ce4
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 3.0.0-SNAPSHOT

We are making backwards-incompatible changes, hence we need to bump
major version.

Change-Id: I1bbd7d55f3a4f1a8cb31e03be9e9215a7bbca4dd
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bug 8771: add bcpkix to the startup bundles

This will avoid bundle refreshes due to wiring in bcpkix.

Change-Id: Ia7ceb119d96c8ffa642e3d427e3357c2b7e750d3
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump scala to 2.12.2

This updates scala runtime to 2.12 and adjusts features accordingly.

Change-Id: I1cffaa9c4e7ef2e615d41e125d4d75a377acf537
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump netty to 4.1.13

http://netty.io/news/2017/03/10/4-0-45-Final-4-1-9-Final.html
http://netty.io/news/2017/04/30/4-0-46-Final-4-1-10-Final.html
http://netty.io/news/2017/05/12/4-0-47-Final-4-1-11-Final.html
http://netty.io/news/2017/06/09/4-0-48-Final-4-1-12-Final.html
http://netty.io/news/2017/07/06/4-0-49-Final-4-1-13-Final.html

Change-Id: I950a06bcb8e70f07803912a6d7c2f55eb67a0394
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Upgrade Jacoco Listeners 2.4 -> 3.8

Change-Id: I951403569556ded05380bb65e5e30addb643dc64
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Remove outdated src/site documentation

see https://lists.opendaylight.org/pipermail/odlparent-dev/2017-July/001218.html

Change-Id: Ie1580e11fa449ed38864960c868c4930f847e291
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Control karaf archive by parameters

- add properties karaf.archiveZip and karaf.archiveTagGz
- by default all archives are being created as before

See also: BUG-8765

Change-Id: I303bc2966d3436b7080d2ac47c7b2c735e537a40
Signed-off-by: Michal Rehak <mirehak@cisco.com>

${project.build.directory} in maven-bundle-plugin

Instead of hard-coding "target/" as directory name.
for the LICENSE file. This fixes the following probem:

mvn -Pq,ide clean install

[INFO] --- maven-bundle-plugin:3.0.1:bundle (default-bundle) @ ... ---
[ERROR] Bundle ... : Input file does not exist: target/classes/LICENSE
[ERROR] Error(s) found in bundle configuration

Locally verified that with this fix, the problem shown above does not
happen anymore - and the LICENSE file is still in bundle JARs.

PS: In order to reproduce the problem above (and test this fix),
you (obviously) need to 'rm -rf target/ target-ide/' first... ;)

Change-Id: I313478ec95c9a6e647a595418bd92d45bbb04a8d
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Bug 8799: populate repo first

To avoid overwriting patched files, we populate the local repository
before overwriting files with contents from
opendaylight-karaf-resources. This ensures that our patched files
survive karaf4-parent processing into projects’ distributions.

Change-Id: I4c00b493d50e56b117ec3c56ac3ffe2fa7814516
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bug-8798: ensure script permissions are correct

Since https://git.opendaylight.org/gerrit/59486 script permissions
aren’t preserved (from a tarball); this patch enforces permissions on
all the shell scripts.

Change-Id: Ie4ed6b31b20302ea925aeefc472e1c09a3295a1d
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Disable SFT on feature repos by default

Testing the features in repos can cause issues when they are
mutually-incompatble. Since we test features individually anyway, we
can safely disable SFT by default. skip.karaf.featureTest can be set
to false for projects which want to re-enable this.

Change-Id: I02ec0094985ccc5e7cc1f404808b156883c25911
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump BC to 1.57

https://www.bouncycastle.org/releasenotes.html

Change-Id: I802837879550de85af80858bd1c62c8028d04f03
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add Bouncy Castle to startup bundles

We rely on the provider being available, wire it into startup
properties to get that going.

Change-Id: Iee3a386a2d7f6e36caaaa03e846511ad4a213bbe
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bug 8677: Do Karaf patch-ing earlier and only once

This changes the application of those patches over Karaf to happen once
during the opendaylight-karaf-resources build (as part of odlparent),
instead of in (each child of) karaf4-parent, as it used to be.

This is better because it:

(a) solves mysterious issues with patch sometimes failing, see
https://lists.opendaylight.org/pipermail/odlparent-dev/2017-June/001136.html

(b) avoids problems for Windows users (where there is no "patch")

(c) makes Karaf local distro builds in projects (a bit) faster.

Change-Id: I16d912c6bc0d60295a787994a602e2753b37bfb2
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Remove unused k4.0.7 patches

Change-Id: I956ad203e1f8d17b5e5da61f44590e789ad705d8
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Bump versions to 2.1.0-SNAPSHOT

Change-Id: I87425ce991c5277f60074121132254104704d813
Signed-off-by: Thanh Ha <thanh.ha@linuxfoundation.org>

Bug 8571: copy src/main/assembly

The contents of src/main/assembly need to be copied to
target/assembly, overwriting any conflicting file from Karaf.
integration/distribution needs this to be able to add custom scripts
and replace configuration files.

(Arguably this should be part of integration/distribution, but it
can’t be — at least not in a maintainable fashion.)

Change-Id: Ifcce82862d350b87853727e2c77140adee164886
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump org.apache.aries.blueprint.core version

From 1.4.2 to latest 1.7.1, because in
org.opendaylight.controller:blueprint version 1.7.1 was already used
anyway, and indeed this appears to be the version used in our current
version of Karaf, and it seems wiser to declare this in only a single
place, to avoid possible version discrepancy problems re. this artifact
in the future. So let's just bump this one, and drop the version in
controller:blueprint in a follow-up change.

Change-Id: Ib332fb88eab11a376e93ea90304e2a22115b3e85
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Remove '4' suffix in karaf[4]-plugin

Change-Id: I05c6241213d761f4764f7b7db8accb2e100d638c
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Drop odl-guava-18

Change-Id: I545bcacb92d7ede5c95a1f004869d3d35f0438ca
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Clean up POMs

The main point of this patch is to remove duplicates in the artifacts
POM. The patch also uses ${project.version} where appropriate.

Change-Id: I0554adb79bb7060bf0d45dbd3349440b7f808a7c
Signed-off-by: Stephen Kitt <skitt@redhat.com>

BUG-7446: switch advertized guava to 22.0

This is the build-time switchover patch to have downstream
pick up guava 22.

Change-Id: Ic547fb93ba718d51e6be6f6400b259beff83b8b8
Signed-off-by: Robert Varga <rovarga@cisco.com>
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Add antlr4 dependency declaration

This is used in yangtools mostly, but will be useful for anyone
dealing with lexers/parsers.

Change-Id: I5e1b232ca7357a6895162427d35736020859f266
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump Xtend 2.11 → 2.12

https://eclipse.org/Xtext/releasenotes.html#/releasenotes/2017/05/25/version-2-12-0

Change-Id: I538fed69d39af0c493d0eda6642601b1a28a3c7d
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump to 2.0.0-SNAPSHOT

Since we’ve introduced breaking changes, we need to bump to
2.0.0-SNAPSHOT. (Thanks to Michael for pointing out the error of my
ways!)

Change-Id: Ic79385b234526ca0ab72180dbe2664351ed680ab
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump git-commit-id-plugin 2.2.1 → 2.2.2

Change-Id: Id73ca8bee3b8b048ad3f277b2bc40dced19c6b37
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Remove '4' suffix in applicable artifact names

During the Karaf 4 migration, while both Karaf versions coexisted it
was necessary to have a different artifactId for Karaf 4 features and
feature repositories produced by odlparent. After Karaf 3 is removed,
this is no longer necessary, and this patch removes the '4' from
applicable directory, artifact and property names.

Change-Id: Ib6446a86a75a1a4b9637e73b71aca640a22807f4
Signed-off-by: Lorand Jakab <lojakab@cisco.com>

Remove all Karaf 3 Artifacts (now Karaf 4, only)

Change-Id: I4e60779869d8d4cb4760c1327f4abf238be98ddd
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Bump PMD plugin

Change-Id: Ica5d0d9e0751ee058ed97d0f06a5704c2223f7ca
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Use groupId and version from parent

To keep things consistent inherit artifact group identifiers and
versions from their parent. Use the same version everywhere.

Note that checkstyle and license bundles are getting a version bump to
1.9.0-SNAPSHOT as a result.

Change-Id: I3640869b59cc1e452155e59d19a77e1abacd5454
Signed-off-by: Thanh Ha <thanh.ha@linuxfoundation.org>
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump logback version

Bumps to latest 1.2.2.

Change-Id: I02a0bf1210197d4b6e59d5ba8eba17e9e5ee42e9
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump odlparent for next dev cycle

Change-Id: Iae69f160b89e59ade8f3efade03f2c18d3ba9806
Signed-off-by: Thanh Ha <thanh.ha@linuxfoundation.org>

Bug 8685: upgrade jolokia from 1.3.3 to 1.3.6

Changes are described in depth here:
https://jolokia.org/changes-report.html

Change-Id: Ifad9f981d893c686616c5b8ee509da997db5ed37
Signed-off-by: Ryan Goulding <ryandgoulding@gmail.com>

Add more pax-exam artifacts

These artifacts are used by downstream to run tests and need to
declare them as dependencies.

Change-Id: I1f41d44d13513f1abe8f3ae9a8663cd3665ef067
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add Jackson dataformat XML

Change-Id: Ic3aa82a376765e0871369fe119b82d21ca80a319
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Update commons-lang3 3.5 → 3.6

Release notes:
http://www.apache.org/dist/commons/lang/RELEASE-NOTES.txt

Change-Id: Ib0b83a16d93600b746e8a7e938d9415912d398ea
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Remove <prerequisites><maven> avoid mvn WARNING

[WARNING] The project ... uses prerequisites which is only intended for
maven-plugin projects but not for non maven-plugin projects. For such
purposes you should use the maven-enforcer-plugin.

See https://maven.apache.org/enforcer/enforcer-rules/requireMavenVersion.html

Change-Id: I196af33c5a4fc80d6720add92171d5de58d2224c
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Remove @Deprecated bundles4test TestBundleDiag

as infrautils.ready does not use this anymore, now

Change-Id: Idd26ebd139d3ddb7ed8c8ef65c8301497f9158d4
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Bump Checkstyle from 6.16 to 7.6.1

This (likely) requires some tweaks in projects enforcing CS;
needs testing using https://wiki.opendaylight.org/view/EvolvingCheckstyle

Motivated by having had to fix
https://git.opendaylight.org/gerrit/#/c/54978/

Change-Id: Ibef1cbb06e8e3d28a44b687562ebfc400d403b32
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Fix aggregator-parent relativePath

Fixes aggregator-parent's incorrect relative path to odlparent-lite.

Change-Id: Ibab245e65e6ce0583c6c58c219a40ac7dd2c3c7a
Signed-off-by: Thanh Ha <thanh.ha@linuxfoundation.org>

Bug 4290 Add missing org.osgi.service.event

Because apparently we need this, always (without it
opendaylight-karaf4-empty does not start anymore, following c/58053).

Change-Id: Icac86b301bda8136448ec45d3567fe728efc7f10
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

BUG-4219: fix karaf CLI reload

Controller's blueprint extenstion is pulling in an optional
dependency of pax-logging, causing it to reload, effectively
restarting everything. Wire osgi-service before starting up
to eliminate this.

Change-Id: Idf0170d805e98362e7f3560db7f166c68a18d19d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove remote repositories

Our ODL Karaf distributions should never download artifacts from
maven.org or repository.springsource.com (we prepare an offline Maven
repo under the system/ directory in our Karaf distribution during
build).

This also adds the missing Apache Aries Quiesce API to system/ to fix
another problem; this may always have gotten downloaded at runtime, but
we never noticed this problem before removing the remote repos (above),
re. "Could not find artifact
org.apache.aries.quiesce:org.apache.aries.quiesce.api:jar:1.0.0"

I've noticed this while trying to work offline on the plane to SFO for
the DDF! ;)

Change-Id: Ib8d1390c463d22130e3dc46d27ac2951a808a8d6
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Relax Checkstyle RightCurly rule for LITERAL_DO

The change [1] try to avoid checkstyle forcing this
"...} while (condition)" statement to be split in two lines by
removing the LITERAL_WHILE from odl_checkstyle, but the correct
token should have been LITERAL_DO.

LITERAL_WHILE avoids forcing this [2]:
while(line !=null)
{ // We do not want this
   process(line);
   line = in.readLine();
}

LITERAL_DO avoids forcing the following [3]:
do
{ // We do not want this
   x = rand.nextInt(10);
} // neither this
while(x < 5);

[1] https://git.opendaylight.org/gerrit/#/c/55264/
[2] http://checkstyle.sourceforge.net/apidocs/com/puppycrawl/tools/checkstyle/api/TokenTypes.html#LITERAL_WHILE
[3] http://checkstyle.sourceforge.net/apidocs/com/puppycrawl/tools/checkstyle/api/TokenTypes.html#LITERAL_DO

Change-Id: I2054e44f140e7b92480da954c45169a31931d683
Signed-off-by: David Suarez <david.suarez.fuentes@ericsson.com>

Better gitignore rule for vim tmp files

Old one just caught the first backup, this catches additional.

Change-Id: I01802aeee8b578c6cac25bcb626e06e3d194098d
Signed-off-by: Daniel Farrell <dfarrell@redhat.com>

Bug 8415 related: Make TestBundleDiag a lib

This is the alternative to c/56848 requested by Robert.

We're changing this so that we can consume bundles-test-lib in the
infrautils.ready implementation as a service for both Karaf v3 and v4.

As a (VERY nice!) side effect, this allows us to remove the duplicated
Java code we had between bundles-test and bundles4-test. That was,
originally, intended to only be like it still is for a short while, but
with the Karaf 4 situation being what it is, this code duplication is
still there - and the respective code has started diverging a bit; so
unifying this in the new bundles-test-lib is great.

Change-Id: I0fd2c954fdc4a6c84505bef51729d15e4ba2888c
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Pin requests to <2.16

The new requests library pulls in a chardet dependency that is in
conflict with coala packages. Pin the version of requests to ensure
we're pulling in a supported version.

Copied change by Thanh.

Change-Id: I3b29f2ab809a1373a9e148120db9f004f3e3fba2
Signed-off-by: Daniel Farrell <dfarrell@redhat.com>

Pin coala to 0.11

coala-bears 0.11 just released 5 hrs from this patch author date. It
breaks the build because we were pinning coala. It seems coala and
coala-bears versions must match exactly so pin both versions for good
measure.

Change-Id: I044b283be027ec04930ff956583b16daa11bc950
Signed-off-by: Thanh Ha <thanh.ha@linuxfoundation.org>

Pin coala version to <0.11

I seems the new 0.11 version breaks us due to dependency conflicts.
Let's pin the version and migrate when we have time to troubleshoot.

Change-Id: I29b31fb12580e54d9a1761e11d3f75998ff7f43e
Signed-off-by: Thanh Ha <thanh.ha@linuxfoundation.org>

Bug 8176: Bump Pax Exam (and Pax URL) versions

to solve NoRouteToHostException on Fedora-based Linux

see https://ops4j1.jira.com/browse/PAXEXAM-740, and
http://blog2.vorburger.ch/2017/04/java-local-rmi-using-jdks-inetaddress.html

I've locally verified this does indeed fix the original problem.

Change-Id: Iaa4a7c74dc887b2105a80604afea498bb7deac60
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

blueprint-maven-plugin file move made conditional

This makes the move of the generated blueprint XML file conditional,
because we should not unconditionally hard code the move of the
blueprint XML files generated by the blueprint-maven-plugin.

This move from OSGI-INF/blueprint/autowire.xml to
org/opendaylight/blueprint should instead be conditional on the presence
of hand written XML in src/main/resources/org/opendaylight/blueprint
(which, typically, uses the ODL extensions).

This is because it should still be possible to use the @Singleton and
@Inject annotations with the XML generation by blueprint-maven-plugin in
bundles which do not depend on the MDSAL (e.g. in infrautils; I actually
have this case & problem in the real world).

Note that the gen. autowire.xml never uses the ODL extensions (unless we
do more work in the future to generate ODL extensions from annotations;
in which case the move should be made conditional on the XML content
actually containing/sing ODL extensions, not location).

Of course, normal ODL instance will always have the ODL controller mdsal
blueprint integration active. But e.g. in ITs (e.g. in infrautils) it
seems "silly" to have to depend on the full mdsal JUST to get the
blueprint integration working JUST because we hard-coded and enforced
the move here.

This change has been tested by running a "mvn clean package" on e.g.
genius/interfacemanager/interfacemanager-impl, which is a typical ODL
bundle with a src/main/resources/org/opendaylight/blueprint/*.xml with
xmlns:odl="http://opendaylight.org/xmlns/blueprint/v1.0.0” as well as
some @Singleton in *.java, and verified that in the resulting JAR as
expected there are x2 XML in org/opendaylight/blueprint, namely the
handwritten say interfacemanager.xml as well as the generated
autowire.xml - both with and without this change; so no regressions.
Likewise, IFF there is no src/main/resources/org/opendaylight/blueprint,
then (only) the autowire.xml stays in OSGI-INF/blueprint (that's new,
and the goal of this change).

Change-Id: Id283078e0547e41caa0a714fe0f0fd70a048966c
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Bug 8483: Fix maven-dependency-plugin to v3.0.1

In the (vain?) hope that this may prevent NPEs; see bug.

Change-Id: I21d355d82d68625d00651336e928fb8c0d2f35f7
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Introduce new aggregator-parent POM

Just to avoid having that <skip>true for maven-deploy-plugin and
maven-install-plugin copy/pasted all over ODL for each *-aggregator.

This also makes changes as appropriate to use it in odlparent itself.

The intention is to start shifting to using this in other projects too.

Having a new parent artifact for all aggregators MAY also come in handy
when I start looking in earnest at our JavaDoc generation.

Q & TODO: Some, but not all, of the typical *-aggregator projects also
have a maven-site profile with maven-javadoc-plugin - should we move
that here as well? I'm hesitant because that sometimes has
<inherited>false .. which I'm not very familiar with - it would not do
anything with <inherited>false in a <parent>, would it? So do we include
it in aggregator-parent, but without <inherited>false ? Or forget about
it and just start with a simple aggregator-parent as is proposed here
for now (and refine JavaDoc related stuff later) ?

Change-Id: I33fba54051b5d3e5fec145d8ae68785bb66c8725
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Bug 8415 related: TestBundleDiag API improved

required for https://git.opendaylight.org/gerrit/#/c/56749/

The implicit throws org.awaitility.core.ConditionTimeoutException and
java.lang.AssertionError were not great.. an AssertionError is special,
and more importantly Awaitility's ConditionTimeoutException is an
implementation detail.

Awaitility is actually bundled as a private OSGi embedded JAR-in-JAR,
which is a bit like shaded fat über JARs with relocation in non-OSGi
Java built by the maven-shade-plugin or maven-assembly-plugin; so it was
wrong to throw an exception from it to a consumer of this utility.

Therefore, this introduces a new SystemStateFailureException instead.

Change-Id: I7b48f5d39258a62c38a3fb8c98a6a35b97360276
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Bump akka to 2.4.18

http://akka.io/news/2017/05/02/akka-2.4.18-released.html
https://github.com/real-logic/Agrona/releases/tag/0.9.5
https://github.com/real-logic/Aeron/releases/tag/1.2.5

Change-Id: I2a19b3b2b1d35dba28e729938d6c579d55930681
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

javax.inject dependencyManagement <optional>true

Because at production run-time in Karaf we do not actually need these
annotations - they are only consumed by the blueprint-maven-plugin to
generate XML. (In component tests with Guice we do need them at
runtime, but then infrautils.inject.guice.testutils enforces
<optional>false via c/56724.)

Thanks to FELIX-2809, resolution:=optional actually gets added to the
OSGi MANIFEST.MF by the maven-bundle-plugin if in the POM a dependency
is marked as <optional>true.

NB that even though javax.inject.Inject & Co. have
@Retention(RetentionPolicy.RUNTIME), this does work - because in Java
annotations can be absent at runtime when loading classes (contrary to
other linked classes; annotations are handled specially, see JLS).

This should be safe & backwards compatible to change, because it just
means we don't have to have javax.inject at runtime anymore, but for
those existing features which include a wrap:mvn:javax.inject already,
it normally should not cause any disruption either.

see also https://ops4j1.jira.com/browse/PAXCDI-230

Change-Id: I0246e5d150fd018a26fd3a9ca46b2c96b430bc94
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Introduce pax-cdi-api into dependencyManagement

This enables @Inject @OsgiService instead of <reference> in XML; see
https://wiki.opendaylight.org/view/BestPractices/DI_Guidelines#.40OsgiService_.26_.40OsgiServiceProvider

see c/56723 for full details re. why it is <optional>true

Change-Id: I1aa5f314cb561dbbd82cbd34c4812e0e1a803a4b
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Add pax-url-link & pax-exam-link-assembly

into dependencyManagement of odlparent's POM.

I'm experimenting with using http://veithen.github.io/alta/ to use Pax
URL's link: protocol and (eventually) completely do away with the mvn:
URLs from pax-url-aether (which is a stupid approach, not by us/ODL but
originally by Pax Exam).

Merging this isn't urgent yet, as I can just put the version into the
POM of where I'm experimenting with this, but ultimately it would be
good to have this in odlparent; just thought I'd raise it early.

Change-Id: Ic372ad5db0db9c26bbfce2f14592408365a1d5d7
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

skip.karaf.featureTest in single-feature-parent

Historically, the skip.karaf.featureTest property could be used to
skip SingleFeatureTest; that was lost in the migration to Karaf 4.
This patch restores the functionality in single-feature-parent.

Change-Id: Id6ed63129d7b9d7284356ca665152b321fc9ca16
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Fix pax-web's commons-collections

pax-web-4.3.0 can still pull in commons-collections-3.2.1, fix that.

Change-Id: Id0e0a678fcc54b45a864b69aed8ba7d1345bb756
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 60cadf07b66747fd5e9823edcb8ef61b1f33b804)

Align pax-web and karaf standard features

Karaf's standard feature uses ow2/asm-all 5.0.4, while pax-web
is installing 5.0.2. Patch pax-web to use the same version, saving
around 230k.

Change-Id: Ifb29d1b3d3a54d4d55818795289b3eb0e357a826
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Use odl4-guava in odl4-akka-leveldb

Our Karaf 4 features should only use other Karaf 4 features.
odl4-akka-leveldb-0.7 ended up pulling odl-guava-18 instead of
odl4-guava-18... This patch fixes that, and changes odl4-guava-18 to a
dependency in odl4-akka-leveldb-0.7.

Change-Id: I1206579b51adaa159e0d629db3fbdbf03fcce455
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Update dependencies for Scala 2.11.11

Change-Id: Ibe0cfd9975d3f8e5d97ae040dba12f5532cce3c1
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump scala to 2.11.11

Scala-2.11.11 is the last release in the 2.11 train, update to it.

Change-Id: I2d053d10b35b55efff0359f4303a66e8099e810b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Simplify feature dependencies

Our Karaf 4 features include a number of dependencies which are also
transitive dependencies, and which don’t constitute the primary
API/runtime we’re interested in providing. This patch removes these
features (the generated features are identical).

Once we remove the corresponding Karaf 3 features we’ll also be able
to remove some of these dependencies from dependency management
altogether.

Change-Id: Ifaa7db8afc1128c331b2cdb710dd95887c55788d
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Allow logging feature dependency changes

This adds two properties controlling the Karaf feature generator’s
tracking features.

Setting checkDependencyChange to true enables checks and logs: when
features are generated, a simplified form saved in src/main/history;
builds where this file is already present compare newly-generated
features against them, and log any failures. Any detected changes in
features are logged as a warning, with the detail of the changed
feature contents. It would be interesting to parse this in the builds
and send emails when changes are detected.

Setting failOnDependencyChange fails the build if a dependency change
is detected in the features.

This patch also enables both these flags in odlparent features where
it’s useful. This ensures that any changes to those features are
caught.

Change-Id: If9a1ebb2b218413f78a1bc28bfeb20d39b956675
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Karaf 4: allow SFT configurations to be combined

We need to allow children of single-feature-parent to reconfigure the
generator plugin, while still ensuring that the list of excluded
artifacts is honoured.

Change-Id: Ifa284e6e10c5182a68d4a3457446903ef9d1ccb8
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Revert "Bug 8208: revert to the Karaf 4.0.7 rt"

This reverts commit dacab8c7e49bc18578ceb8f088333550cfe389ab.

Change-Id: I6c7808b1ef130287a52c6d24c20c951f7d5dcd53
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit d3fa42cfd8347ef5763a3aacb0a5e1ad325f5d7e)

Clean up lambdas

Change-Id: Iedee023ba95d010aa71886b8946c0562358e4979
Signed-off-by: Stephen Kitt <skitt@redhat.com>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bug 7900: make feature repos aggregate features

Instead of redefining features and producing standalone feature
repositories, feature-repo-parent can produce features which list the
individual feature repositories produced by single-feature-parent.
This allows integration/distribution's integration-index to build much
faster (by rendering SFT moot there).

SFT now tests individual features in single-feature-test instead of
testing the full repository. This makes the -Psft profile useless so
it is removed. It also means that odl-bundles-test needs to be handled
specially, since it is used in features4-test: it can't use
single-feature-parent, otherwise it introduces a dependency loop. This
patch introduces a new, internal parent,
untested-single-feature-parent, which is used by odl-bundles-test and
single-feature-parent: untested- takes care of the actual feature
construction, and single-feature-parent adds the SFT tests.

Change-Id: Ib0de6ce1e6488944bcbe0aa7e84496676f6db19a
Signed-off-by: Stephen Kitt <skitt@redhat.com>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Replace explicit types with <>

Change-Id: I1933cd82a1917955d8c363c47727e9a4dc895cef
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bug 8261: prevent TestBundleDiag from throwing NPE

There is a stack trace demonstrating this behavior in bug 8261.  In
short, TestBundleDiag is occasionally throwing NPE because a null
Bundle is returned from serviceReference.getBundle().  This change just
adds some null sanity checking in TestBundleDiag and
ServiceReferenceUtil.

Also, in ServiceReferenceUtil, a local variable is extracted just in
case the object is mutated between calls.  This at least gives a
consistent snapshot for usingBundles over the course of the method
call.

Change-Id: I334b0850be4a2045f9eeb260f5de4af542896dc1
Signed-off-by: Ryan Goulding <ryandgoulding@gmail.com>

Bug 7926: exclude OSGi bundles from features

We need to avoid having OSGi core bundles pulled in by features (they
cause bundle refreshes which break Pax Exam tests). This patch marks
the OSGi bundles as provided, and excludes them from generated
features.

Change-Id: I66f519e4d498f956cb3fa087d98d30a64b3316f4
Signed-off-by: Stephen Kitt <skitt@redhat.com>

BUG-8208: remove prereq from jdbc/jetty features

It seems that dependencies have to be completely converged
and prerequisite=true does not play nice with optional imports,
leading to half-baked installation.

This patch drops dependency/prerequisite flags, which will need
to be revisited once we have this thing working.

Change-Id: I9340e7dff6566b793689406e8029b546ec349302
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Relax Checkstyle RightCurly rule for LITERAL_WHILE

In the Checkstyle upgrade from 6.16 to 7.6.1 done in parallel in other
changes, I've come across a number of "violations" because it did not
like "... } while (condition)" in one line.

This indeed seems to be what our current Checkstyle rule would mandate,
but because of what looks like a bug in our currently used version 6.16
was never enforced. As we upgrade, they seem to have fixed that, so
either we add a \n after } before while during the upgrade in the
projects which are already enforcing Checkstyle - or we just relax this
rule - personally I think it's ugly on two lines, and thus suggest this.

See http://checkstyle.sourceforge.net/config_blocks.html#RightCurly

I have not actually really tested this, just seems above right.

Change-Id: I21e20d91b854340a636c4fbbe5a6038caf0bee57
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

BUG-8202: make war a prerequisite

This seems to help getting the container up and about in local testing,
but it may break something else.

Change-Id: I0cadae70de7ef39001fccfaa33982de3293cc130
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit e9f860ea8c2c9a01b4e9e4f12e8ece04e62e91df)

Remove akka-distributed-data-experimental

This module is causing trouble with org.objectweb.asm and is not
used anywhere anymore. Remove it.

Change-Id: I0dcb1d73533f2a7ef1999093840c0c93a4f1450d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bug 8208: revert to the Karaf 4.0.7 runtime

We keep the 4.0.9 Maven plugin to get the new generator features, but
switch back to the 4.0.7 runtime to avoid the resolver issues.

Change-Id: Ifebe6ab8155a86c2e9e8c80315ae345a7f033a79
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump versions by x.(y+1).z for next dev cycle

Change-Id: Ie775e9ef6b719033459b6b61bfae7c464b0f2a6f
Signed-off-by: Anil Belur <abelur@linuxfoundation.org>

Bug 4219: upgrade to Karaf 4.0.9

Change-Id: Ibfaf2b64c971ab4320d8114cb74d4f5e4c2ae305
Signed-off-by: Stephen Kitt <skitt@redhat.com>

BUG-8117: fix jetty/pax-web configuration mismatch

Karaf 4 uses pax-web as the jetty configuration bridge, which
compares the jetty configuration with its directives. If a port
or address mismatch is found, this leads to duplicate ports being
used and hence preventing successful jetty startup.

Change-Id: Ia5b71621ba0474481dab738232ebac778540a258
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Drop karaf-maven-plugin

We can’t use it within odlparent anyway, and the release of Karaf
4.0.9 means we don’t need it.

Change-Id: I04edebf6799115d35b225e145d18dc29235819a1
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Add .tox/ to .gitignore

Change-Id: I002f31717abcb15ed5e4d0b91ea774b6fc79203f
Signed-off-by: Michael Vorburger <vorburger@redhat.com>

Bug 8175: Improve SFT TestBundleDiag under Karaf 4

This does NOT yet completely fix all problems of the SingleFeatureTest
TestBundleDiag under Karaf 4 (which was in really bad shape!), most
importantly there is still something somewhere in Pax Exam that I
haven't understood yet which seems to just "swallow" all exceptions
thrown by SingleFeatureTest's @Test installFeature() but at least now:

1. there is something in the log now for failures at least
thanks to a (try/catch/log) WORKAROUND (!)

2. LinkageError fixed (see details below)

3. IllegalStateException: BundleContext is no longer valid fixed (?)

This changes the setup by just embedding bundles4-test and it's
dependencies into features4-test, instead of programmatically installing
the odl-bundles-test feature (which is removed in this change) or the
features4-test bundle.

This simplifies installation, avoid problems with the wrap: protocol,
and should resolve weird problems we've seen (only!) in the distribution
jobs, which seem to have had something to do with a custom local Maven
repo.

IMHO it would be *REALLY* good to practice proper TDD for SFT (with
TestBundleDiag) now, and have some automated negative testing of broken
features and bundles (incl. BP) in odlparent...

Bug: 7981 related later proper solution
Change-Id: I516c0f6c4e9ee85f0c753cb4b58a0b53fc22c7ea
Signed-off-by: Michael Vorburger <vorburger@redhat.com>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>