Flatpak External Data Checker

This is a tool for checking for outdated or broken links of external data in Flatpak manifests.

Motivation

Flatpak apps are built using external data (git checkouts, tarballs,
simple files, etc.). A very specific case of this is the use of extra
data, which works as a way to download third party binaries at
installation time.

Of course, the links pointing to external data can become obsolete, so
it is very important to account for and correct such issues.
This is especially critical for the extra data, in which a broken link
impedes the installation of the application.

This project offers ways to easily check or monitor the state of such
links, as well as the suggestion of new versions.

It works by extracting all the external data of a Flatpak manifest and
giving it to a collection of checkers, which will set up the right
state and, possibly, new versions of each external data.

Use

The simplest use of this tool is by calling:

flatpak-external-data-checker MANIFEST_FILE

it should display messages about any broken or outdated external data.

Installation

This tool itself is available in flatpak format from Flathub. Install with

flatpak install --from https://dl.flathub.org/repo/appstream/org.flathub.flatpak-external-data-checker.flatpakref

And run with

flatpak run org.flathub.flatpak-external-data-checker MANIFEST_FILE

Running in a container

flatpak-external-data-checker is also avaiable as an
OCI image
from GitHub Container Registry.

You can use the run-in-container.sh helper script to set up needed CLI
options for you and run the image using podman:

~/src/endlessm/flatpak-external-data-checker/run-in-container.sh \
    [ARGS …] \
    ~/src/flathub/com.example.App/com.example.App.json

Automatically submitting PRs

When run with the --update flag, this tool can commit any necessary changes
to Git and open a GitHub pull request. In order to do this, it requires
a GitHub access token,
specified in the GITHUB_TOKEN environment variable.

Automatically merging PRs

The tool will also automatically merge previously opened pull request for
unavailable (BROKEN) sources if the change has successfully passed CI checks
and the token has sufficient privileges.

Automatically merging all submitted PRs, not just unavailable sources,
from the tool can be forcefully enabled by setting
automerge-flathubbot-prs to true in flathub.json,
or it can be completely disabled by setting automerge-flathubbot-prs to false.

Changes to Flatpak manifests

For simple checks to see if a URL is broken, no changes are needed. However,
you can add additional metadata to the manifest to allow the checker to
discover new versions.

Some of the following checkers are able to determine upstream version number,
and automatically add it to releases list in metainfo. To specify which source
is the app upstream source, set property is-main-source to true in the
checker metadata for that source.

URL checker

If the upstream vendor has an URL that redirects to the latest version of the
application, you can add something like the following to check and update the URL for
the latest version:

"x-checker-data": {
    "type": "rotating-url",
    "url": "http://example.com/last-version",
    "pattern": "http://example.com/foo-v([0-9.]+).tar.gz"
}

The version number for the latest version can be detected in two ways:

  • If the filename ends with .AppImage, the version number is extracted
    from the AppImage. (It is run in a bwrap sandbox.)
  • Otherwise, if "pattern" is specified in "x-checker-data", the given
    regular expression is matched against the full
    URL for the latest version, and the first match group is taken to be the
    version. (This follows the convention used by
    debian/watch files.)

HTML checker

Both the version number and the download URL will be gathered from a static
HTML page which contains this information:

"x-checker-data": {
    "type": "html",
    "url": "https://www.example.com/download.html",
    "version-pattern": "The latest version is ([\\d\\.-]+)",
    "url-pattern": "(https://www.example.com/pub/foo/v([\\d\\.-]+)/foo.tar.gz)"
}

The HTML checker also supports building the download URL using
the retrieved version string, its components according to the Python
LooseVersion
class and semantic versioning fields:

"x-checker-data": {
    "type": "html",
    "url": "https://www.example.com/download.html",
    "version-pattern": "The latest version is ([\\d\\.-]*)",
    "url-template": "https://www.example.com/$version/v$version.tar.gz"
}

"x-checker-data": {
    "type": "html",
    "url": "https://www.example.com/download.html",
    "version-pattern": "The latest version is ([\\d\\.-]*)",
    "url-template": "https://www.example.com/$major.$minor/v$version.tar.gz"
}

If the placeholder is immediately followed by an underscore, you need to add braces:

"x-checker-data": {
    "type": "html",
    "url": "https://www.example.com/download.html",
    "version-pattern": "The latest version is ([\\d\\.-]*)",
    "url-template": "https://www.example.com/$version0.$version1/v${version0}_${version1}_version2.tar.gz"
}

If the HTML page contains multiple versions with download links,
set single pattern containing two nested match groups for both url and version:

"x-checker-data": {
    "type": "html",
    "url": "https://sourceforge.net/projects/qrupdate/rss",
    "pattern": "<link>(https://sourceforge.net/.+/qrupdate-([\\d\\.]+\\d).tar.gz)/download</link>"
}

To disable sorting and get first matched version/url, set sort-matches to false.

Git checker

To check for latest git tag in corresponding git source repo, add checker
metadata with type git and set tag-pattern to a regular expression with
exactly one match group (the pattern will be used to extract version from tag):

"x-checker-data": {
    "type": "git",
    "tag-pattern": "^v([\\d.]+)$"
}

By default tags are sorted based on version number extracted from tag.
To disable sorting and keep order from git ls-remote, set sort-tags to false.

If the project follows semver specification, you can set
version-scheme property to semantic in order to use semantic version scheme for sorting.
In this case, make sure that tag-pattern extracts only valid semver strings.

JSON checker

The JSON checker allows using jq to query
JSON data with arbitrary schema to get version and download url.

To use the JSONChecker, specify JSON data URL, version query and url query
(you can use $version variable got from the version query in url query):

{
    "type": "json",
    "url": "https://api.github.com/repos/stedolan/jq/releases/latest",
    "version-query": ".tag_name | sub(\"^jq-\"; \"\")",
    "url-query": ".assets[] | select(.name==\"jq-\" + $version + \".tar.gz\") | .browser_download_url"
}

for git type sources, specify tag query and, optionaly, commit and version queries:

{
    "type": "json",
    "url": "https://api.github.com/repos/stedolan/jq/releases/latest",
    "tag-query": ".tag_name",
    "version-query": "$tag | sub(\"^jq-\"; \"\")",
    "timestamp-query": ".published_at"
}

timestamp-query is optional, but if provided - must return a string with timestamp in ISO format.

See the jq manual for complete information about writing queries.

Debian repo checker

For the DebianRepoChecker, which deals only with deb packages, it
can read the following metadata (add it to manifest element it refers
to, e.g. where "type": "extra-data" is declared):

"x-checker-data": {
    "type": "debian-repo",
    "package-name": "YOUR_PACKAGE_NAME",
    "root": "ROOT_URL_TO_THE_DEBIAN_REPO",
    "dist": "DEBIAN_DIST",
    "component": "DEBIAN_COMPONENT"
}

Anitya (release-monitoring) checker

Anitya is an upstream release monitoring
project by Fedora. It supports multiple backends for retrieving version information
from different services, including GitHub, GitLab, Sourceforge, etc.
To use the AnityaChecker, specify numeric project ID on release-monitoring.org
and add a template for source download URL. This checker supports the same placeholders
as the HTMLChecker:

"x-checker-data": {
    "type": "anitya",
    "project-id": 6377,
    "stable-only": false,
    "url-template": "https://github.com/flatpak/flatpak/releases/download/$version/flatpak-$version.tar.xz"
}

Set stable-only to true to retrieve latest stable version (as recognized by Anitya).

For git type sources, instead of url-template, set tag-template to derive git tag from version.

GNOME checker

Check for latest source tarball for a GNOME project.

"x-checker-data": {
    "type": "gnome",
    "name": "pygobject",
    "versions": {
        "<": "3.38.0"
    },
    "stable-only": true
}

Set stable-only to false to check for pre-releases, too.

You can also set version constraints in versions property (optional).
It should contain key-value pairs where key is the comparison operator
(one of <, >, <=, >=, ==, !=), and the value is the version to compare with.
So, {"<": "3.38.0", "!=": "3.37.1"} means "any version less than 3.38.0 except 3.37.1".
All constraints must match simultaneously, i.e. if one doesn't match - version is rejected.

PyPI checker

Check for Python package updates on PyPI.

"x-checker-data": {
    "type": "pypi",
    "name": "Pillow"
}

By default it will check for source package (sdist package type).
To check for binary package instead, set packagetype to bdist_wheel
(only noarch wheels are supported currently).

JetBrains checker

Special checker that will check for available updates
for JetBrains products:

"x-checker-data": {
    "type": "jetbrains",
    "code": "PRODUCT-CODE",
    "release-type": "release or eap (defaults to release)"
}

Snapcraft checker

Special checker that will check for available updates
for Snapcraft packages:

"x-checker-data": {
    "type": "snapcraft",
    "name": "PACKAGE-NAME",
    "channel": "stable, beta, or any other tag the project uses"
}

Rust checker

Special checker that will check for available updates
for Rust:

"x-checker-data": {
    "type": "rust",
    "package": "package name, for example: rust",
    "channel": "nightly, stable or beta",
    "target": "target triple, for example: x86_64-unknown-linux-gnu"
}

Chromium checker

Special checker that will check for available updates to the
Chromium tarballs, as well as the toolchain
binaries or sources used to build it.

"x-checker-data": {
    "type": "chromium",
    "component": "chromium, llvm-prebuilt, or llvm-git; defaults to chromium"
}

The following components are supported:

  • chromium: updates the Chromium tarball itself, used on URL-based sources
    (e.g. type: archive).
  • llvm-prebuilt: updates a tarball pointing to the official LLVM prebuilt
    toolchain archive matching the latest Chromium version. Used on URL-based
    sources.
  • llvm-git: updates a type: git source for its commit to point to the LLVM
    sources for the toolchain used by the latest Chromium version.

GitHub

https://github.com/flathub/flatpak-external-data-checker