grim/convey
closing merged branch
Convey is a pipeline for running containers and protecting the host as much as possible. It accomplishes this goal by creating a volume and making it available to each item in the pipeline.
Convey was created out of the need to build [Pidgin](https://pidgin.im) and its supporting libraries on multiple Linux distributions, Windows and MacOS. While there are many continuous integration tools out there, all of them still require maintenance and can only be used by the project itself.
This becomes a problem when you have a community of plugin developers who would also like to build against all of its supported operating systems but can't and shouldn't need to put forth the effort and maintenance involved with using other available tools.
This is where Convey's concept of [build containers](#markdown-header-concepts) comes into use. Once you have all the dependencies required to build something, like a Pidgin plugin, you are then able to share that Docker image with everyone using this system and they are then able to use that image to build against. So a project like Pidgin can supply build images containing the most recent releases with all of the build dependencies for multiple platforms and then plugin developers can use that image to test their plugins against.
This also works when it comes to deployments. You can create an image to deploy to [bintray](https://bintray.com/) so that users are able to specify options in their convey.yml and not have to worry about the specifics of how it's uploaded.
If you still have doubts about how Convey can simplify your workflow and increase productivity, please contact me via [email](mailto:grim@reaperworld.com) or by opening an issue and let me know!
Convey is build around a concept that I call a build container. A build container is a container that has all of the dependencies to turn an input into something else. This could be source code or even just data processing. For example the `convey/go-build` image is a simple wrapper around the `golang` image that just builds a project instead of installing it. You can find more information on build containers [here](http://blog.terranillius.com/post/docker_builder_pattern/).
Convey is built around this concept to allow you to build/process anything on any platform while at the same time protecting the host from a malicious config. This protection is provided by not exposing dangerous options like volume mounts or port forwards while running containers. This is accomplished by creating a volume a data volume at the start of the build plan and volume mounting it into containers as they are run. This allows you to pass state between modular tasks that can be reused between multiple plans.
Once built, convey can build itself entirely. This is done by using an import task to copy the source code into the workspace. Once that stage is complete, it runs another stage that builds for linux-x86_64, windows-x86_64, and darwin-x86_64 in parallel. The final stage will take those compiled binaries and export them back to the hosts file system.
Of course you don't have to export back to the host file system. Since these are just containers, you could instead upload them to your staging environment, packagecloud.io, bintray.com, whatever. Of you could export to the host and create another plan that will import the artifacts and then publish them.
Convey supports multiple configuration loader which can be specified by the `-l` command line argument.
Convey supports multiple different types of configs that it can load. Currently supported loaders are convey and bitbucket.
The `bitbucket` config is for [bitbucket-pipelines](https://bitbucket.org/product/features/pipelines) which will allow you to run your pipeline locally. Convey does a best attempt at emulating the pipelines environment, but your mileage may vary. If you're having problems, please open an issue.
To use the bitbucket config loader, just run `convey -l bitbucket`. If there is a `bitbucket-pipelines.yml` file in the working directory, convey will happily run it.
bitbucket pipelines does not have plans in the same sense that convey does, so if you do not specify one, convey will use the same heuristics that bitbucket-pipelines does to run the correct pipeline from the config.