Skip to content

Latest commit

 

History

History
344 lines (234 loc) · 10.7 KB

development.md

File metadata and controls

344 lines (234 loc) · 10.7 KB

Development

Environment setup

Enable oVirt packages for Fedora:

sudo dnf copr enable -y nsoffer/ioprocess-preview
sudo dnf copr enable -y nsoffer/ovirt-imageio-preview

Enable virt-preview repository to obtain latest qemu and libvirt versions:

sudo dnf copr enable @virtmaint-sig/virt-preview

Update the system after enabling all repositories:

sudo dnf update -y

Fork the project on https://github.com/oVirt/vdsm.

Clone your fork:

sudo dnf install -y git
git clone [email protected]:{your_username}/vdsm.git

Install additional packages for Fedora, CentOS, and RHEL:

contrib/install-pkg.sh

Generate the Makefile (and configure script):

./autogen.sh --system --enable-timestamp

Now you can create the virtual environment (https://docs.python.org/3/library/venv.html), which is necessary to run the tests later. This needs to be done only once:

make venv

Building Vdsm

Before building, it is recommended to recreate the Makefile because it contains version numbers, which might have changed by updating the local repository:

./autogen.sh --system --enable-timestamp

To build Vdsm:

make

To create the RPMs:

make rpm

To upgrade your system with local build's RPM (before you do this you should activate maintenance mode for Vdsm):

make upgrade

Running the tests

To run tests, first enter the virtual environment:

source ~/.venv/vdsm/bin/activate

Then start some tests with tox, for example the networking tests:

tox -e network

To exit the virtual environment afterwards:

deactivate

For more information about testing see /tests/README.md.

Making new releases

Release process of Vdsm version VERSION consists of the following steps:

  • Changing Version: field value in vdsm.spec.in to VERSION.

  • Updating %changelog line in vdsm.spec.in to the current date, the committer, and VERSION.

  • Committing these changes, with subject "New release: VERSION" and posting the patch to GitHub.

  • Verifying the patch by checking that the GitHub CI build produced a correct set of rpm's with the correct version.

  • Merging the patch (no review needed).

  • Tagging the commit immediately after merge with an annotated tag: git tag -a vVERSION

  • Making a new release in the GitHub repo.

CI

Running tests locally is convenient, but before your changes can be merged, we need to test them on all supported distributions and architectures.

When you push patches to GitHub, CI will run its tests according to the configuration in the .github/workflows/ci.yml file.

Advanced Configuration

Before running make you could use ./configure to set some (rarely used) options. To see the list of options: ./configure -h.

SPDX headers

All files must include the SPDX copyright notice and the license identifier. This project employs reuse to handle copyright notices and ensure that all files have the proper SPDX headers. To add the SPDX headers to new files in the project you can use:

contrib/add-spdx-header.sh new_file.py

This will create default GPL-2.0-or-later license header with Red Hat, Inc. as copyright holder.

# SPDX-FileCopyrightText: Red Hat, Inc.
# SPDX-License-Identifier: GPL-2.0-or-later

To add new license to be used in the project:

reuse download <License-Identifier>

Check list of available license identifier in https://spdx.org/licenses/.

To add SPDX header to a file with a non-default license:

reuse addheader
  --copyright="Red Hat, Inc." \
  --license="<License-Identifier>" \
  --template=vdsm.jinja2 \
  --exclude-year \
  new_file.py

Please check that all files are reuse-compliant before pushing your branch:

make reuse

Coding guidelines

The coding conventions in Vdsm are generally aligned with the PEP 8 style recommendations. Code is linted with flake8 and pylint based in the /.pylintrc file. You can trigger the linters check by pushing changes into your fork, which invokes:

make lint

However, some code conventions are not explicitly enforced by the linters due to their limitations. Nonetheless, new code shall adhere to those rules, and old code should be updated as possible when patched.

In general, it is very important for new code to be consistent with the existing code in the same module, or in the same package for new modules. In case of doubt, check the already existing code before!

In this section, we will cover some of the most important conventions and preferences for contributing to Vdsm, as extension to PEP 8 general rules.

Imports

First import standard library modules, then 3rd party modules, and finally project modules. Separate standard library modules and 3rd party modules with a blank line. Keep alphabetical order within each block.

Less specific imports:

from vdsm.foo import baz

Should come before more specific imports like:

from vdsm.foo.bar import baz

Relative name imports are accepted, but better to avoid.

Indentation

Use 4 spaces per indentation level.

Continuation lines should align wrapped elements either vertically using Python's implicit line joining inside parentheses, brackets and braces, or using a hanging indent. When using a hanging indent, the following should be considered:

  • There should be no arguments on the first line.
  • Further indentation should be used to clearly distinguish itself as a continuation line.
  • Hanging indentation should stick to the 4-space rule.
# Add 4 spaces (an extra level of indentation) to distinguish arguments from the rest.
# Consider that having too many arguments in a function is an anti-pattern.
foo = long_function_name(
    var_one, var_two, var_three, var_four)

# Recommended extra indentation with multiline if-statements.
if (condition_1 and condition_2 [...]
        and condition_n):
    do_something()

# But this is also possible.
def long_function_name(
        var_one, var_two, var_three,
        var_four):
    print(var_one)

To break comprehension Lists, Dicts, Sets as well as generator expressions, keep expressions that would be written in a single line together. Continuation lines stick to the same rules as above. Closing brace/bracket/parenthesis, if placed on a separate line, shall line up under the first character of the line start.

# Vertically aligned and extra 4 spaces for condition continuation
foo = [var for var in bar_list
       if condition_1
           and condition_2
           and condition_3]

Docstrings

Always use the three double-quote """ format for docstrings (per PEP 257). A docstring shall be organized as multiple summary lines (one physical line not exceeding 79 characters) terminated by a period. First and last line of the docstring shall be blank.

Module functions and classes should have a docstring, unless:

  • Not public.
  • Pertain to a test module.

Class methods do not require docstring, but is recommended when it is a public entity or have a complex interface.

The docstring should be imperative-style (e.g., Verify configuration file.) rather than descriptive-style (e.g., Verifies configuration file.). The docstring should describe the function's main purpose, its calling syntax, and/or semantics, but avoid implementation details.

Certain aspects of a function shall be documented in special sections, listed below.

Args: (or Arguments:)

List each parameter by name. It is strongly encouraged that the type of the parameter follows the name, either enclosed in parentheses, with clear wording, or in type annotation format. Then, the description of the parameter separated by a colon. If the type is optional or can take multiple types, that can be specified with comma-separated lists or use type annotation syntax. If the description of the parameter is too long to fit in 79 characters, following hanging lines can be indented with 2 or 4 extra spaces (as long as it is consistent with surrounding docstrings).

Returns: (or Yields: for generators)

Start with the type of the return value, followed by a colon and its semantics. If the function only returns None, this section is not required. It may also be omitted if the docstring starts with Return or Yield (e.g., Return a boolean indicating if the folder has the right permissions.), and the opening sentence is sufficient to describe the return value.

Raises:

List all exceptions that are relevant to the interface, followed by a description. Use a similar exception name + (type) + colon + space and hanging indent style as described in Args:.

def example_function(arg_1, arg_2=None):
    """
    This is an example docstring. Start with summary explanation of the
    function main purpose. Can expand multiple lines.

    Args:
        arg_1 (_type_): Long description can be split into
            multiple lines with extra indentation.
        arg_2 (_type_): _description_

    Returns:
        _type_: _description_

    Raises:
        OSError: _description_
    """

Comment blocks and inline comments

Comments are accepted in non-obvious parts of the code that need extra explanation. A comment can expand through multiple lines before the operation starts. Comments should have normal punctuation and grammar. Start sentences with uppercase letters, end with periods, and use commas where necessary. Comments should never describe the code.

Inline comments are allowed, but discouraged. Inline comments are acceptable for very short comments, but shall never expand through multiple lines. To improve legibility, multiple inline comments in successive lines should be vertically aligned.

    _MULTIPATHD = cmdutils.CommandPath(
        "multipathd",
        "/usr/sbin/multipathd",  # Fedora, EL7
        "/sbin/multipathd")      # Ubuntu

    # List of multipath devices that should never be handled by Vdsm. The
    # main use case is to filter out the multipath devices the host is
    # booting from when configuring hypervisor to boot from SAN. This device
    # must have a special rule to queue I/O when all paths have failed, and
    # accessing it in Vdsm commands may hang Vdsm.

Git commits

We cultivate a discipline of clean, linear git history that dates back to gerrit days. Commit messages matter - we have a template for them. Every commit should be atomic and releasable. We don't use merge commits - we use rebasing.