Workflow for the Fortran stdlib contributors

This document describes our current workflow.

We welcome everyone and anyone to participate and propose additions to stdlib. It is okay if you do not have experience for specification or implementation, but have an idea for stdlib. If the idea is popular among the community, more experienced contributors will help it through all 5 steps.

1. Idea: You have an idea or a proposal. Open an issue to discuss it. This is on the level of "is there interest in having image reader/writer functions in stdlib?" The goal of this step is to find out if the community is interested in having this functionality as part of stdlib.

2. API: When there seems to be significant interest in the proposal (vast majority of participants think it is a good idea), move on to discuss the specific API. It's OK to propose the API off the bat if you already have an idea for it. This step is exploratory and its goal is to find out what the API should look and feel like.

3. Specification: Discuss the API and iterate. When there is vast majority approval for the API, move on to implement it and submit a PR. Small PRs are always better than large. It is OK to implement only a few functions of a new module, and continue work on the others in a later PR. All new functionality goes into an "experimental" namespace (version: experimental). As part of the PR, when submitting a new public facing API, please provide the initial draft of the specification document as well as the initial reference implementation of this specification. The specification is a document that describes the API and the functionality, so that anyone can use it to create an implementation from scratch without looking at stdlib. The stdlib library then provides the reference implementation.

4. Implementation in experimental: When opening a PR, request reviews from one or more people that are most relevant to it. These are likely to be people involved in prior steps of the workflow. Other contributors (not explicitly invited) are encouraged to provide reviews and suggestions as well. Iterate until all (or most) participants are on the same page. A merge is permitted if there are unit tests for a majority of the possible calling scenarios (with or without optional arguments, with arguments that trigger an error) and if there is vast majority approval of the PR.

5. Release: Moving from experimental to release. The experimental "namespace" contains new functionality together with its specification. In order to move from experimental to release, the specification document must be approved by the wide community and the standards committee (informally). If that happens, it has now been blessed for broad use and we can move the code into the main section of stdlib, and the particular specification document becomes part of the Fortran Standard Library.

Note: the general term "vast majority" above means at least 80%, but ultimately it is left to our best judgement to ensure that the community agrees that each PR and proposal was approved by "vast majority".

You are welcome to propose changes to this workflow by opening an issue.

Build systems

This project supports two build systems, fpm and CMake.

CMake build files

The build files for CMake allow both in-tree, i.e. build artifacts share the same tree as the source files, and out-of-tree builds, i.e. build artifacts exist in a separate directory tree. Both build types are explicitly supported and tested, the latter strategy is recommended for local development.

Sources for the main library target are added in src/CMakeLists.txt relative to the library target, i.e. no absolute paths are required.

To add tests, the macro ADDTEST should be used instead of the CMake function add_test, the macro hides creation of the executable target, linking against the main library target and registering the test. The tests themselves are defined as standalone executables in the subdirectories in test, a new subdirectory with tests has to be registered in test/CMakeLists.txt.

The source tree should be considered read-only. References to PROJECT_SOURCE_DIR and CMAKE_CURRENT_SOURCE_DIR should only be used for accessing source files, never to write build outputs, use PROJECT_BINARY_DIR and CMAKE_CURRENT_BINARY_DIR to write build artifacts instead. To fully support in-tree builds, build artifacts must never have the same name as source files to avoid accidentally overwriting them, e.g. when preprocessing or configuring a file.

The CMAKE_INSTALL_PREFIX should only be written to on install, never in the build process. To install generated files, create a build output in the build tree and install it with the install function. This project follows the GNU install conventions, this means that the variables CMAKE_INSTALL_BINDIR, CMAKE_INSTALL_LIBDIR, and CMAKE_INSTALL_INCLUDEDIR must be used instead of bin, lib, and include, respectively. Library targets should be exported on install to allow correct inclusion of the project in other CMake projects. Prefer dashes as in project-config or project-targets over camel-case as in projectConfig or projectTarget for file names as the former allows easier construction from the PROJECT_NAME variable by concatenation.

The project is usable as CMake subproject. Explicit references to CMAKE_SOURCE_DIR and CMAKE_BINARY_DIR must be avoided to not break subproject builds. An example project is available here to test the CMake subproject integration.