Skip to content

Latest commit

 

History

History
371 lines (301 loc) · 12.8 KB

README.md

File metadata and controls

371 lines (301 loc) · 12.8 KB

Continuous Integration Travis

libmodulemd

C Library for manipulating module metadata files

Full details can be found in the API Documentation

Using libmodulemd from Python

Using libmodulemd from Python is possible thanks to the gobject-introspection project. To use libmodulemd from Python, include the following at the top of your sources.

Install the python2-libmodulemd or python3-libmodulemd package for your system depending on the version of python with which you are working. These packages provide the appropriate language bindings.

import gi
gi.require_version("Modulemd", "2.0")
from gi.repository import Modulemd

Working with repodata (DNF use-case)

The libmodulemd API provides a number of convenience tools for interacting with repodata (that is, streams of YAML that contains information on multiple streams, default data and translations). The documentation will use two repositories, called "fedora" and "updates" for demonstrative purposes. It will assume that the content of the YAML module metadata from those two repositories have been loaded into string variables fedora_yaml and updates_yaml, respectively.

Tools such as DNF that are consuming data from a repository should always set strict=False, so that it can safely handle minor, backwards-compatible changes to the modulemd format.

First step is to load the metadata from these two repositories into ModulemdModuleIndex objects. This is done as follows:

C

ModulemdModuleIndex *fedora_index = modulemd_module_index_new ();
gboolean ret = modulemd_module_index_update_from_string (
  fedora_index, fedora_yaml, FALSE, &failures, &error);

ModulemdModuleIndex *updates_index = modulemd_module_index_new ();
gboolean ret2 = modulemd_module_index_update_from_string (
  updates_index, updates_yaml, FALSE, &failures, &error);

Python

fedora_index = Modulemd.ModuleIndex.new()
ret, failures = fedora_index.update_from_string(fedora_yaml, False)

updates_index = Modulemd.ModuleIndex.new()
ret, failures = updates_index.update_from_string(updates_yaml, False)

The failures are a list of subdocuments in the YAML that failed parsing, along with the reason they failed. Hence, by checking the return value of failures we will know if the YAML parsing was successful or not.

Since it doesn't really make sense to view the contents from separate repositories in isolation (in most cases), the next step is to merge the two indexes into a combined one:

C

ModulemdModuleIndexMerger *merger = modulemd_module_index_merger_new ();

modulemd_module_index_merger_associate_index (merger, fedora_index, 0);
modulemd_module_index_merger_associate_index (merger, updates_index, 0);

ModulemdModuleIndex *merged_index =
  modulemd_module_index_merger_resolve (merger, &error);

Python

merger = Modulemd.ModuleIndexMerger.new()

merger.associate_index(fedora_index, 0)
merger.associate_index(updates_index, 0)

merged_index = merger.resolve()

At this point, you now have either a complete view of the merged repodata, or else have received an error describing why the merge was unable to complete successfully. Additionally, it should be noted that the combined metadata in any ModulemdModuleIndex will have all of its component parts upgraded to match the highest version of those objects seen. So for example if the repodata has a mix of v1 and v2 ModulemdModuleStream objects, the index will contain only v2 objects (with the v1 objects automatically upgraded internally).

Now, we can start operating on the retrieved data. This guide will give only a brief overview of the most common operations. See the API specification for a full list of information that can be retrieved.

Discover the default stream for a particular module.

C

ModulemdModule *module =
  modulemd_module_index_get_module (merged_index, "modulename");
ModulemdDefaults *defaults = modulemd_module_get_defaults (module);
if (defaults)
  {
    printf ("Default stream for modulename is %s\n",
            modulemd_defaults_v1_get_default_stream (
              MODULEMD_DEFAULTS_V1 (defaults), NULL));
  }

Python

module = merged_index.get_module("modulename")
defaults = module.get_defaults()
print("Default stream for modulename is %s" % defaults.get_default_stream())

Get the list of RPMs defining the public API for a particular module NSVCA

First, query the ModulemdModuleIndex for the module with a given name.

C

ModulemdModule *module =
  modulemd_module_index_get_module (merged_index, "modulename");

Python

module = merged_index.get_module("modulename")

Then, query the ModulemdModule for the ModulemdModuleStream associated with the provided NSVCA (name-stream-version-context-architecture identifier).

C

ModulemdModuleStream *stream = modulemd_module_get_stream_by_NSVCA (
  module, "modulestream", 0, "deadbeef", "coolarch", &error);

Python

stream = module.get_stream_by_NSVCA("modulestream", 0, "deadbeef", "coolarch")

Lastly, read the RPM API from the ModulemdModuleStream. Here, api_list is a list of strings containing package names.

C

GStrv api_list = modulemd_module_stream_v2_get_rpm_api_as_strv (
  MODULEMD_MODULE_STREAM_V2 (stream));

Python

api_list = stream.get_rpm_api()

Also note that in addition to accessor API methods, many objects also have properties that can be accessed directly.

C

printf ("Documentation for module stream is at %s\n",
        modulemd_module_stream_v2_get_documentation (
          MODULEMD_MODULE_STREAM_V2 (stream)));
g_autofree gchar *doc;
g_object_get (MODULEMD_MODULE_STREAM_V2 (stream), "documentation", &doc, NULL);
printf ("Documentation for module stream is at %s\n", doc);

Python

print("Documentation for module stream is at %s" % stream.get_documentation())
print("Documentation for module stream is at %s" % stream.props.documentation)

Retrieve the modular runtime dependencies for a particular module NSVCA

C

ModulemdModule *module =
  modulemd_module_index_get_module (merged_index, "modulename");
ModulemdModuleStream *stream = modulemd_module_get_stream_by_NSVCA (
  module, "modulestream", 0, "deadbeef", "coolarch", &error);
GPtrArray *deps_list = modulemd_module_stream_v2_get_dependencies (
  MODULEMD_MODULE_STREAM_V2 (stream));

for (gint i = 0; i < deps_list->len; i++)
  {
    GStrv depmodules_list =
     modulemd_dependencies_get_runtime_modules_as_strv (
       g_ptr_array_index (deps_list, i));

    for (gint j = 0; j < g_strv_length (depmodules_list); j++)
      {
        GStrv depstreams_list =
          modulemd_dependencies_get_runtime_streams_as_strv (
            g_ptr_array_index (deps_list, i), depmodules_list[j]);

        for (gint k = 0; k < g_strv_length (depstreams_list); k++)
          {
            // do stuff with depmodules_list[j], depstreams_list[k]
          }
      }
  }

Python

module = merged_index.get_module("modulename")
stream = module.get_stream_by_NSVCA("modulestream", 0, "deadbeef", "coolarch")
deps_list = stream.get_dependencies()
for dep in deps_list:
    depmodules_list = dep.get_runtime_modules()
    for depmod in depmodules_list:
        depstream_list = dep.get_runtime_streams(depmod)
        for depstream in depstream_list:
            # do stuff with depmod, depstream

Working with a single module stream (Packager/MBS use-case)

One limitation of the ModulemdModuleIndex format is that it requires that all module streams loaded into it have both a name and a stream name. This however is not possible when dealing with streams such as a packager would be using (since the build-system auto-generates the module name and stream name from the git repository information. In this case, we need to work with a single module stream document at a time. For this, we will use the ModulemdModuleStream interface.

This example will assume that the module name and stream name have already been determined from the repodata and that they are stored in string variables named module_name and stream_name, respectively.

Python

stream = Modulemd.ModuleStream.read_file(
    "/path/to/module_name.yaml", True, module_name, stream_name
)
v2_stream = stream.upgrade(Modulemd.ModuleStreamVersionEnum.TWO)
v2_stream.validate()

In the example above, we upgraded the stream to v2, in case we were reading from v1 metadata. This will allow us to avoid having to manage multiple code-paths and support only the latest we understand. After that, it calls validate() to ensure that the content that was read in was valid both syntactically and referentially.

Also available is Modulemd.ModuleStreamVersionEnum.LATEST which will always represent the highest-supported version of the ModulemdModuleStream metadata format. This may change at any time.

Getting started with developing

Forking and cloning the sources

The libmodulemd project follows the Github Fork-and-Pull model of development. To get started, create a fork of the upstream libmodulemd sources, clone those locally and create branches on your fork to make changes. When they are ready for review or feedback, create a pull-request.

Prerequisites

Preparing a build environment

Create a container for building the libmodulemd sources. Run the handy setup script in the root of the checkout.

./setup_dev_container.sh [<Fedora Release>]

If unspecified, it will default to "Fedora Rawhide". To select a different Fedora release for the base image, add the release number as an argument to the command. For example:

./setup_dev_container.sh 33

This will automatically pull down a Fedora base image, install all of the packages needed for development and testing and then provide you with a shell inside this environment.

Building the sources

Projects built with the meson build-system require a separate build directory from the source path. The meson command will generate this directory for you.

meson --buildtype=debug -Db_coverage=true debugbuild

The above command (run from the root of the source checkout) will create a new subdirectory - debugbuild - configured to compile with debug symbols and gcov symbols to measure test coverage.

To build the sources, chdir() into the debugbuild directory and run

ninja

To build and run the in-tree tests, use

ninja test

To generate HTML documentation, you can run

ninja modulemd-2.0-doc

(Be aware that the GLib documentation module in meson has some strange quirks and won't recognize newly-added pages without deleting and re-creating the build directory first.)

Running more advanced test suites

In addition to the basic ninja test set of tests, libmodulemd also has a suite of tests performed in the Github Actions environment. These are also containerized and can be run locally on the host. (You will not be able to run them from within the development container shell, as nested containers are not supported.)

To run the container-based tests, you can run the following from the source root:

./.ci/ci-fedora.sh

(Optionally, you can pass the release number as an argument to switch to building and testing against that release rather than Fedora Rawhide).

Support for running the tests on other OSes is ongoing. See the .ci directory for available suites. All supported OSes and release versions are tested as part of each pull request.

Tips and tricks

Running tests in debug mode

The libmodulemd library is built atop GObject. It provides a debug mode that is configurable by an environment variable. In general, it is highly recommended that you run all tests with G_DEBUG='fatal-warnings,fatal-criticals' set in the environment. This will cause the application to abort() on programming errors that would be logged and ignored at runtime.

Running tests with valgrind

Assuming your current working directory is debugbuild as described above:

meson test --suite=ci_valgrind --wrap=../contrib/valgrind/valgrind_wrapper.sh

If not, you may need to adjust the path to libmodulemd-python.supp.

You can also specify individual tests to run against. See meson test --list for the available tests.

The automated CI tests will always run with valgrind on all platforms where it is supported.

Authors: