Hi Chris,
thanks fort he changes.
I checked the PR and all artefact seem to be as they should be.
So are there any other issues to be addressed before spinning up the next round
(RC3)?
Julian
Am 05.08.19, 14:46 schrieb "Christofer Dutz" <christofer.d...@c-ware.de>:
Hi all,
I just submitted a PR: https://github.com/apache/incubator-iotdb/pull/332
So I invested a great portion of Sunday and most of today to clean up the
build a bit. Here comes a summary to what I found and changed:
* Nothing is built outside the target directory anymore
* I moved the static parts of the client and server distributions to
“src/assembly/resources” and added an assembly that builds the client and
server inside the target directory (You need to update the documentation on
this)
* The sever is now available at
server/target/iotdb-server-0.9.0-SNAPSHOT
* The client is now available at
client/target/iotdb-client-0.9.0-SNAPSHOT
* I updated the integration tests to work with these new locations
* I updated the distribution module (binary-distribution) to simply
include and unpack the client and server assemblies, which greatly simplifies
the assembly itself.
* In the integration-tests you use a process builder to get the current
working directory … you can access this in the “user.dir” System property.
* There was a large mixture of artifacts with different versions in the
classpath.
* I set all external dependencies to the same versions using a big
dependencyManagement block in the master pom
* I removed the explicit version of every external dependency and
moved it into the main dependencyManagement block in the root pom (except
dependencies only used in the examples, for these I added a
dependencyManagement block in the example pom). This way you instantly identify
external dependencies.
* I added an enforcer rule to fail the build if two differing
versions of one artifact are being used in the build (This should eliminate
this problem from re-occuring)
* I replaced the usage of “${project.version}” with the real version
(There is no need to manually update these versions and this way when looking
at the pom you instantly know the version … keep in mind if an artifact is
deployed to maven central, the user doesn’t have the code and therefore he
always has to scroll up to the top in order to see which version is used)
* There were some imports related to Jdk 11 no longer providing some
APIs, I moved those imports to a maven profile, which is activated based on the
JDK version
* HDFSInputTest created a file in “spark/test/” … if the test failed,
the rat plugin would complain the next time it’s run, I let it create the file
in the “target” directory
However I did notice that there is quite a number of artifacts still in the
classpath that have different maven coordinates, but very similar content:
* com.fasterxml.jackson.* and org.codehaus.jackson (1.9.13) (Not
dangerous as they have different package names)
* javax.activation:activation (1.1.1) and
javax.activation:javax.activation-api (1.2.0) (Probably not dangerous as the
API versions are pretty similar, but the package names are equal)
* javax.servlet:javax.servlet-api (3.1.0) and javax.servlet:servlet-api
(2.5) (Possibly dangerous as there is a difference in the major version, which
usually implies incompatible changes)
* …
* I added an enforcer rule, that reports this type of problems, but
commented it out as there is quite a bit of this throughout the project which
would have to be cleaned up first.
I checked the full build with all tests on Oracle-Java 8, Open-JDK 11,
Oracle-JDK 12.
Chris
