Object

This is a very simple dependency handling system. It's based on naming conventions and plain archive files.

Normally, you should use thinks like conan to handle dependencies. But, if for some obscur reason this is not an option, then this can ease the pain.

A plain software build is relying on

1. source code (the one you write)
2. libraries that needs to be identified and versionned

A software build might in turn produce a library used by yourself or someone else. The following figure illustrate the point.

                    <source code>
                          A
                          |
                    [cmake build] -> <package>.tar.gz -> [artifact-copy] -> /your/repository
                          |
                          V
/your/repository  <-[artifact-deploy]

This module comes as commands

• artifact-copy : move packages into your repository in safe way
• artifact-deploy : deploy packages listed in a requirment file into your workspace.

How to use it

The dependency handler can be build and packaged via this command:

$ python3 setup.py sdist
running sdist
running egg_info
...
copying artifacts.egg-info/top_level.txt -> artifacts-0.1/artifacts.egg-info
Writing artifacts-0.1/setup.cfg
Creating tar archive
removing 'artifacts-0.1' (and everything under it)

This package is installed with the following command:

$ pip install ./dist/artifacts-0.1.tar.gz
Processing ./dist/artifacts-0.1.tar.gz
Requirement already satisfied: PyYAML==5.1.2 in ./venv/lib/python3.7/site-packages (from artifacts==0.1) (5.1.2)
Requirement already satisfied: semantic-version==2.8.2 in ./venv/lib/python3.7/site-packages (from artifacts==0.1) (2.8.2)
Installing collected packages: artifacts
  Running setup.py install for artifacts: started
    Running setup.py install for artifacts: finished with status 'done'
Successfully installed artifacts-0.1

How it's done

Requirements are listed in a file in which the deploy command expects to find a root entry requires.

# This is a sample requirement files
requires:
  stable:
    - openssl-qnx-2.2.3
    - cpp-pthread-qnx-3.2.3
    - double-qnx-1.0.0
  snapshot:
    - snapshot-lib-qnx-1.2.3

The command artifact-deploy uses such a yaml file to list the packages your project's requires. For each entry found, the program expects to find an archive in one of the pathes listed by the variable PACKAGES-HOME_PATH. A package archive name is expected to have this form: <name>[-<os>]-<semver>[-snapshot]-<target arch>.tar.gz.

Where

• name: speaks for itself
• os: the operating system the package was built for (Darwin, ...) - optional
• semver: semantic version (more on semver here)
• snapshot: build type (either snapshot or stable)
• target arch: target CPU (x86, armv7, ...)

When artifact-deploy finds a package, a digest is calculated and checked against the content of a digest file (using MD5).

WARN a digest file is expected to be named <name>[-<os>]-<semver>[-snapshot]-<target arch>.tar.gz.md5 and MUST exist.

$ artifact-deploy --install-dir /your/workspace --target-arch x86 requirements.yml 
deploy artifacts found in ['requirements.yml'] here /tmp/deps
installed 'artifact-qnx-2.3.4-snapshot' found here '/your/repsitory' for 'x86', here '/your/workspace'
artifact-deploy deployed 1 artifacts.
$

When you have built new packages, you can copy them to your repository directory this way.

$ artifact-copy cmake-build/*.tar.gz /your/repository
copied cmake-build/artifact-qnx-2.3.4-snapshot-x86.tar.gz to /your/repository/.
package cmake-build/cpp-pthread-Darwin-1.11.0-x86.tar.gz is stable and exists in /your/repository/, it will NOT be copied.

How you can help

More soon...