If `/boot` is an automount, then the unit will be stopped as soon as the
automount expires. That's would defeat the purpose of using systemd to
delay finalizing the deployment until shutdown. This is not uncommon as
`systemd-gpt-auto-generator` will create an automount unit for `/boot`
when it's the EFI System Partition and there's no fstab entry.
To ensure that systemd doesn't stop the service early when the `/boot`
automount expires, introduce a new unit that holds `/boot` open until
it's sent `SIGTERM`. This uses a new `--hold` option for
`finalize-staged` that loads but doesn't lock the sysroot. A separate
unit is used since we want the process to remain active throughout the
finalization run in `ExecStop`. That wouldn't work if it was specified
in `ExecStart` in the same unit since it would be killed before the
`ExecStop` action was run.
Fixes: #2543
It can be useful to parse the options and initialize the sysroot without
actually loading it until later. Factor out the sysroot loading to a new
`ostree_admin_sysroot_load` and add a new
`OSTREE_ADMIN_BUILTIN_FLAG_NO_LOAD` flag to accommodate this.
Make a copy of `man/html` to `docs/man` and then configure Jekyll to
include it verbatim like the API docs. A link is added to the main index
and the necessary commands are added to the github docs workflow.
Add a --enable-man-html configure option to build HTML man pages using a
different stylesheet. The HTML pages aren't installed as I don't know
what purpose they'd serve on an actual installation.
I really don't know what these were for. In both man and HTML output
with a recent version of the Docbook XSL, the command synopses are
already indented. I don't see any reason why they'd need to be manually
indented more.
Since Ubuntu 18.04, libgpgme-dev is the real package and libgpgme11-dev
is a virtual package provided by it. Apparently LGTM running on Ubuntu
20.04 no longer resolves the virtual package:
```
WARNING: Package 'libgpgme11-dev' requested by configuration file was not found
```
That ends up causing the build to fail:
```
configure: error: Need GPGME_PTHREAD version 1.1.8 or later
```
The commit metadata `version` key is well established but getting it for
a remote commit is cumbersome since the commit object needs to be
fetched and loaded. Including it in the summary additional metadata
allows a much more convenient view of what each of the remote refs
represents.
The ubuntu-latest VMs are currently based on 20.04 (focal). In focal,
libseccomp2 doesn't know about the close_range syscall[1], but
g_spawn_sync in impish tries to use close_range since it's defined in
glibc. That causes libseccomp2 to return EPERM as it does for any
unknown syscalls. g_spawn_sync carries on silently instead of falling
back to other means of setting CLOEXEC on open FDs. Eventually it causes
some tests to hang since once side of a pipe is never closed. Remove
this when libseccomp2 in focal is updated or glib in impish handles the
EPERM better.
1. https://bugs.launchpad.net/ubuntu/+source/libseccomp/+bug/1944436Fixes: #2495
As pointed out in the original review, `gpg-list-keys` fits better
alongside the existing `gpg-import`.
Changes were done with:
```
git grep -l list-gpg-keys | xargs sed -i 's/list-gpg-keys/gpg-list-keys/'
for src in $(git ls-files '*list-gpg-keys*'); do
dst=${src/list-gpg-keys/gpg-list-keys}
git mv "$src" "$dst"
done
```
Currently the verifier decides whether to include the global keyrings
based on whether the specified remote has its own keyring or not. Allow
callers to exclude the global keyrings even when that's not the case.
This will be used in a subsequent commit in order to get the GPG keys
only associated with a remote.
Rather than use the release codename tags, use the release stage tags.
This way the configuration (theoretically) doesn't need to be updated
when new Debian and Ubuntu releases are made.
For Debian stable is used instead of buster and a testing (bullseye)
build is added. For Ubuntu, latest is used instead of focal for the
current LTS and rolling is used instead of groovy for the latest
release. This actually changes the Ubuntu build from groovy to hirsute.
This refreshes the build dependencies installed for the GitHub Tests
workflow based on the Build-Depends in the upstream packaging. The
handling is now more explicit about any deviations and any release
differences.
Don't cancel all the jobs if one distro config fails. The jobs are
mostly independent, so we do want to let the others continue in case
the failure is isolated to that particular distro configuration.
travis-ci.org stopped running builds on June 15, 2021. Since this
organization is very unlikely to switch to travis-ci.com, just drop the
setup. The new GitHub Actions tests completely replace it.
This runs the test suite in various distros. The intention is to use
this to replace the Travis CI setup since it often has rate limit
failures.
Each configuration in the matrix runs in a Docker container, installs
system dependencies and then builds and tests ostree. The scripts are
basically copy and paste of the travis ones with some of the lesser used
features pruned out.
Some differences from the travis setup:
* OS details are gathered from `/etc/os-release` instead of being passed
in as environment variables.
* The scripts always assume the user is root and don't try to use
`sudo`.
* The `installcheck` test has been removed since ostree doesn't actually
use that. It could be added to run the installed tests or
`gnome-desktop-testing-runner` could just be called directly.
There should be enough flexibility to run other distros like Fedora,
Arch or Alpine. Another option would be to use the other build scripts
in ci/.
In configure the systemd unit path is optional, but in the code it's
assumed to be defined. Add an `#ifdef` that throws an error when it's
not defined like the handling of `HAVE_LIBMOUNT` below it.
The semantics of multiple process locking are covered by
test-concurrency.py, but the semantics of the repository locking from a
single process aren't handled there.
This checks how the repository locking is handled from a single thread
with one OstreeRepo, a single thread with multiple OstreeRepos, and
multiple threads sharing an OstreeRepo.
If there's a locking issue in this test, then it's likely not going to
resolve after a few seconds of serializing access. Lower the default 30
second lock timeout to 5 seconds to prevent the test from hanging
unnecessarily.
Use `g_error` and `g_assert*` rather than `g_return*` when checking the
locking preconditions so that failures result in the program
terminating. Since this code is protecting filesystem data, we'd rather
crash than delete or corrupt data unexpectedly.
`g_error` is used when the error is due to the caller requesting an
invalid transition like attempting to pop a lock type that hasn't been
taken. It also provides a semi-useful message about what happened.
Previously each thread maintained its own lock file descriptor
regardless of whether the thread was using the same `OstreeRepo` as
another thread. This was very safe but it made certain multithreaded
procedures difficult. For example, if a main thread took an exclusive
lock and then spawned worker threads, it would deadlock if one of the
worker threads tried to acquire the lock.
This moves the file descriptor from thread local storage to the
`OstreeRepo` structure so that threads using the same `OstreeRepo` can
share the lock. A mutex guards against threads altering the lock state
concurrently.
Fixes: #2344
This simplifies the lock state management considerably since the
previously pushed type doesn't need to be tracked. Instead, 2 counters
are kept to track how many times each lock type has been pushed. When
the number of exclusive locks drops to 0, the lock transitions back to
shared.
In gnupg 2.3.0[1], if a primary key is expired and a subkey does not
have an expiration or its expiration is older than the primary key, the
subkey's expiration will be reported as the primary's. Previously a
subkey without an expiration would not report one regardless of the
primary key's expiration.
This caused a regression in a test setting an expiration on a primary
key. The test was checking that the subkey was not expired by asserting
that there was no `Key expired` line in the signature verification
output. With gnupg 2.3.0+, it will show as expired, causing the test to
fail.
Remove the assertion since it's not consistent across gnupg versions. In
practice we don't care whether the subkey is considered expired or not
as long as the signature verification fails when the primary key is
expired.
1. https://dev.gnupg.org/T3343Fixes: #2359
The ostree repo has read permissions set for workflows, which prevents
the documentation job from pushing the built docs to the gh-pages
branch. Raise the job's permissions to write for repo contents to allow
that.
Make a copy of `apidoc/html` to `docs/reference` and then tell Jekyll to
include it verbatim. This will include the gtk-doc API docs on the
static site. A link is added to the main index.
A script is added to do the copy (a symlink won't do) and is setup to
run before Jekyll in the GitHub workflow. Ideally this would be a local
Jekyll plugin to make the process automatic, but the github-pages gem
doesn't allow that.
This uses the Jekyll Actions GitHub action to push the rendered docs to
the gh-pages branch rather than GitHub's automated docs flow. That will
allow greater control over how the docs are generated. Pushing to the
gh-pages branch only happens on pushes to main. For pull requests, the
docs are only built.
This mimics the GitHub Pages environment so that you can build and serve
the site locally for testing. It's will also be required later for using
Jekyll Actions[1] instead of the automated GitHub Pages flow.
1. https://github.com/marketplace/actions/jekyll-actions
This returns a 404 since the site is already generated from the docs
directory. Furthermore, the `CONTRIBUTING.md` markdown file isn't in the
generated site, just the HTML.
Instead, use jekyll's `link` tag to create the link. Unfortunately,
before jekyll 4.0 (github-pages uses 3.9), you have to prepend the base
URL.
Several tests generate summaries and then expect to use the generated
summary immediately. However, this can cause intermittent test failures
when they inadvertantly get a cached summary file. This typically
happens when the test is run on a filesystem that doesn't support user
extended attributes. In that case, the caching code can only use the
last modified time, which only has 1 second granularity. If tests don't
carefully manage the summary modification times or the repo cache then
they are likely subject to races in some test environments.
This introduces an environment variable `OSTREE_SKIP_CACHE` that
prevents the repo from using a cache directory. This is enabled by
default in tests and disabled for tests that are a explicitly trying to
test the caching behavior.
Fixes: #2313Fixes: #2351
Although people are likely not deploying on 32 bit x86 anymore,
deploying on 32 bit armv7 is still often used. Add back an i386 build on
debian to try to catch 32 bit bugs in CI.
Currently the Ubuntu builds are stuck waiting for an answer on what
timezone to use. That could be fixed, but generally the way to do these
types of installs is to set the DEBIAN_FRONTEND environment variable to
`noninteractive` so that debconf just chooses a default. This is what
debootstrap does, for instance. I tested installing tzdata on a local
focal container this way and it just chooses `Etc/UTC` as the timezone.
If you don't configure with --enable-gtk-doc, these won't be included in
the generated .gitignore. If you later configure with --enable-gtk-doc
in the same checkout, they'll show up as untracked files. Include them
in GITIGNOREFILES to ensure they're always added regardless of configure
options.
Like the top level .gitignore, there's no reason to track this if it's
being generated automatically. It only produces spurious diff changes
that you almost certainly don't want.
The local pull path was erroring on any missing commit, but that
prevents a depth pull where the source repo has truncated history. As in
the remote case, this also tries to pull in a tombstone commit if the
source repo supports it.
Fixes: #2266
When pulling with depth, missing parent commits are ignored. However,
the check was applying to any commit, which means that it would succeed
even if the requested commit was missing. This might happen on a
corrupted remote repo or when using ref data from a stale summary.
To achieve this, the semantics of the `commit_to_depth` hash table is
changed slightly to only ever includes parent commits. This makes it
easy to detect when a parent commit is being referenced (although there
is a minor bug there when multiple refs are being pulled) while keeping
references to commits that need their `commitpartial` files cleaned up.
It also means that the table is only populated on depth pulls, which
saves some memory and processing in the common depth=0 case.
Fixes: #2265
In some cases such as backups or mirroring you may want to pull commits
from one repo to another even if there commits that have incorrect
bindings. Fixing the commits in the source repository to have correct
bindings may not be feasible, so provide a pull option to disable
verification.
For Endless we have several repositories that predate collection IDs and
ref bindings. Later these repositories gained collection IDs to support
the features they provide and ref bindings as the ostree tooling was
upgraded. These repositories contain released commits that were valid to
the clients they were targeting at the time. Correcting the bindings is
not really an option as it would mean invalidating the repository
history.
Dan Nicholson