I am setting up an ant build system on a project with dependency resolution being managed by ivy. I have it up and running with the file system being used for the local and shared repository currently. My ultimate goal would be that when developers are fixing bugs or creating new functionality, they would only be able to put artifacts into their local repository. When they belive their code is ready to be used by the rest of the team, it would be promoted to the proper branch in SVN and the group in charge of doing official builds would compile and publish the new artifacts.
So I guess my questions are how can you control who can publish to a repository? Does ivy just rely on filesystem permissions?
Also, I would eventually like to make my shared repository available via http. I think I could point apache to the file system repository directories for retrieving artifacts, but how do you setup publishing to an http repository?
I would suggest that you setup a repository manager to manage your project's build artifacts.
The best choices are one of the following:
nexus
artifactory
archiva
Publishing to a Maven repository means that your artifacts can be consumed by projects using other build technologies. All modern build systems support Maven (Including ivy, see the ibiblio resolver).
You could specify three resolvers in your ivy settings file. First would be a chain resolver which include remote and local ivy repositories. Second would be a local resolver for local ivy repository. Third a resolver to remote ivy repository only.
Every developer retrieves artifacts using first chain resolver.
Usual developer publish artifacts using second local resolver.
Your special team could use third remote resolver to publish in remoter ivy repository.
To protect remote repository from usual developers place it on (S)FTP server with write protection by password.
The only problem in this case is how to set versions on artifacts so that artifacts published in remote repository in some cases override locally published in some not.
Our team used such scheme few years ago. But now we use only local ivy repositories and CI server to build and run tests from various branches. We came up to this after switching to git.
for existing ivy repo easy to setup this: rest-ivy
Related
I want to create a local maven2 repository so we can download dependencies faster.
we use sbt
what is the simplest way to take all or some depedencies from http://oss.sonatype.org/content/repositories/releases and put it in our server
I was thinking of a script to download some jars and put it in a sftp server and use sftp resolver with sbt
https://github.com/harrah/xsbt/wiki/Resolvers
A pretty simple way would be to install a repository manager. A popular one is Sonatype Nexus (the application that serves http://oss.sonatype.org). The repository manager is installed on your server, and set up to proxy online Maven repositories. Then you download your dependencies from the repository manager as though it was the actual server. The first time you download a dependency, it'll be as slow as before (the proxy will download it from the internet), but after that, it'll serve its cached copy. There are some details on this question.
I think there should be some difference, but can anyone tell me the details?
mvn:install copies your packaged Maven module to your local repository (by default, in ~/.m2/repository), to be accessed by other local Maven builds.
mvn:deploy uploads your packaged Maven module to another (usually remote) repository, to be accessed by other, not necessarily local, Maven builds.
See the documentation for the build lifecycle for more info.
The install phase is responsible for the installation of artifacts into local caching repositories. This basically applies to the Maven repository, but a well-known example is also the OSGi Bundle Repository supported by maven-bundle-plugin.
The deploy phase is responsible for the installation of artifacts into published repositories. This usually applies to remote repositories, but is could perfectly be a local repository exposed to the outside world.
As all Maven phases, you can do with them anything you want. You can shuffle plugin phases as you see fit, but the above semantics is the conventional one and you should stick to it in order to be consistent with the default phases of other plugins' goals.
mvn:deploy performs deployment to remote reposiory/environment, mvn:install installs all compiled packages to a local repository making them available to other builds performed on the local machine.
In one sentence: mvn:install compiles and installs your component in your local Maven repository, so that you can use it when other components used and developed locally depend on it. mvn:deploy deploys your (previously installed) component to a remote repository.
We are using SVN as a Maven Repositary. We knew that it isn't advisable to have SVN backed maven repository, but still we have to go with it due to limitation within our organization. SVN backed repository used currently is partially implemented.We have developers working from two different geographical location. Problem we face now is whenever a developer adds a artifact to maven repo(svn repository) all other developers have to update the local svn view manually to get the newly added artifact before we do mvn clean package.
IS there a way to automatically download the artifact from svn maven repository to local repo if the artifact doesn't exists locally?
SVN is hosted with a webserver so maven repositary is accessed using HTTPS protocol only. We use maven 2.2 version.
I tried with wagon plugin which would deploy the build output(jar\war) to scm directly. We are not interested in deploying the build outputs. We need a solution to download artifacts automatically from svn maven repo if it isn't exists locally?
You make no mention of how your SVN repository exposes it's artifacts to the development teams. If it truly a Maven repository (conforming to the standard Maven repository layout) then you could just specify it's URL in the "repositories" section of your POM. Updating the local repo would then no longer be necessary.
I suspect that what you have checked into subversion is not a Maven repository layout? You'd lose one of the key benefits of using subversion if each new version of an artifact was being checked in as a new file....
You are describing the functionality offered by any Maven repository manager, for example: Nexus.
I understand your reluctance to embrace a new repository technology, but SCM systems like subversion are primarily designed for tracking changes to textual files.
In conclusion, if you truly wish to keep subversion in the loop I'd suggest one of two options:
Use subversion to control the contents of the local repository. (3rd party dependencies and the artifacts generated by the developers)
Use a repository manager like Nexus. Let Nexus manage cached content from external repositories, but commit the contents of locally hosted repositories into Subversion.
I have open-sourced some of my projects. I'd like to publish the artifacts in a common place. If I don't have to host my own repository, that would be ideal, otherwise I will have to setup my own instance of Nexus (or other repository that you recommend).
Can I publish artifacts to Maven Central? What process do I need to follow to get them there?
Yes you can, especially if it is an open source project. Here is some info:
http://central.sonatype.org/pages/ossrh-guide.html
My team uses an internal team maven repo that is shared from a development server using Apache. We also run the Continuum CI server on the same machine. Maven builds in Continuum are run with the "install" goal, which copies the final artifact directly into the shared directory.
The question is, what is the difference between adding files to the shared repo using mvn install and using the deploy goal (mvn-deploy plugin)?
It seems to me that using mvn deploy creates additional configuration hassles, but I have read somewhere that installing files into a shared repo is a bad idea for some reason related to the internal workings of maven.
update: I get the functional differences between deploy and install; I am actually more interested in the low level details in terms of what files are created in the maven repo.
Ken, good question. I should be more explicit in the The Definitive Guide about the difference. "install" and "deploy" serve two different purposes in a build. "install" refers to the process of installing an artifact in your local repository. "deploy" refers to the process of deploying an artifact to a remote repository.
Example:
When I run a large multi-module project on a my machine, I'm going to usually run "mvn install". This is going to install all of the generated binary software artifacts (usually JARs) in my local repository. Then when I build individual modules in the build, Maven is going to retrieve the dependencies from the local repository.
When it comes time to deploy snapshots or releases, I'm going to run "mvn deploy". Running this is going to attempt to deploy the files to a remote repository or server. Usually I'm going to be deploying to a repository manager such as Nexus
It is true that running "deploy" is going to require some extra configuration, you are going to have to supply a distributionManagement section in your POM.
From the Maven docs, sounds like it's just a difference in which repository you install the package into:
install - install the package into the local repository, for use as a dependency in other projects locally
deploy - done in an integration or release environment, copies the final package to the remote repository for sharing with other developers and projects.
Maybe there is some confusion in that "install" to the CI server installs it to it's local repository, which then you as a user are sharing?
"matt b" has it right, but to be specific, the "install" goal copies your built target to the local repository on your file system; useful for small changes across projects not currently meant for the full group.
The "deploy" goal uploads it to your shared repository for when your work is finished, and then can be shared by other people who require it for their project.
In your case, it seems that "install" is used to make the management of the deployment easier since CI's local repo is the shared repo. If CI was on another box, it would have to use the "deploy" goal.