I have a cache drive I often mount read-only, and the previous commit
for opening `tmp/cache` broke since `errno == EROFS`, not `EPERM`.
It turns out we already had the concept of a "writable" repo, so just
piggy back off that.
Closes: #281
Approved by: giuseppe
To GLnxConsoleRef. There were some subtleties here, for example we
used to reference `GSConsole` inside the progress changed function,
which at first seems like an ABI hazard, because e.g. rpm-ostree or
xdg-app could still be passing a `GSConsole` instance there. Luckily,
it turns out to be compatible to just start calling libglnx here.
Another issue was that due to libglnx's use of the cleanup function,
we needed to ensure we always called `ostree_async_progress_finish()`
*before* the cleanup function was invoked.
Closes: #280
Approved by: giuseppe
It's actually just easier for build systems (e.g. rpm-ostree)
using dracut to use `--add ostree` rather than indirecting
through the conf file.
This makes it easier for yum-managed systems to install ostree without
side effects.
Closes: #279
Approved by: gatispaeglis
There was some leftover intermediate cruft here I noticed
while reviewing another patch:
- We had an output `GFile*` for that was never used
- We required the caller to allocate the loose pathbuf, but
none of them ever reused it
- We had an extra intermediate function
Also while looking at this, I'm now uncertain whether some of the
callers of `_ostree_repo_has_loose_object` should really be invoking
`ostree_repo_has_object()`, but let's leave that aside for now.
Closes: #272
Approved by: alexlarsson
I'm not sure if something changed or there was a race condition , or
if extraterrestrials intervened to make the tests pass before, but
in the Docker container `/dev/fuse` does exist and since we appear
to be running as uid 0 it'll be writable.
We previously discussed that another requirement for FUSE (and most
setuid binaries) is having `CAP_SYS_ADMIN`, so let's check for that in
the bounding set too.
Closes: #266
Approved by: jlebon
We keep forgetting to update `apidoc/ostree-sections.txt`, so let's
start enforcing it. Of course it turns out we had some bugs here
like symbols marked as public but never implemented, etc. Those
are fixed in the prior commits.
Closes: #263
Approved by: giuseppe
This was removed in 14d682305b - long
before we had a stable library.
I again claim removing unimplemented symbols from the header is not an
API/ABI break.
Closes: #263
Approved by: giuseppe
If a symbol falls in a git merkle tree forest, but no one hears it,
did it ever exist?
I'm claiming this isn't an API/ABI break because nothing could
have actually used this. `git log -S checksum_update_meta` leads
me to 38ef75e6e0 which was before
we really had a stable shared library.
Closes: #263
Approved by: giuseppe
It's not quite namespaced enough to have
`/usr/libexec/grub2-15_ostree`, and the Fedora spec file expects
things in `/usr/libexec/ostree`. Changing the spec file would be
annoying as we'd need conditionals.
Closes: #249
Approved by: gatispaeglis
I think it's cleaner if we use `remote_repo_local` to know
that we have a local repo. In reality, it might be nicest
if we didn't even create an `OstreeFetcher` for this case,
but untangling the code is tricky.
Closes: #239
Approved by: alexlarsson
If one happens to not have grub2 installed, the previous commit
would cause us to fall back to `grub-mkconfig` with bad results.
We should likely just do dynamic detection and avoid pain, but
there's no harm in also allowing this to be statically determined.
The automagic here is that if `/etc/os-release` says `ID_LIKE=fedora`,
we know to use `grub2-mkconfig`. But distro scripts should set this
in the build configurations using `--with-grub2-mkconfig=`.
Closes: #240
Approved by: jlebon
It's not working for me in `make check` on a RHEL 7 Workstation,
apparently because no GPG agent is spawned. I'm guessing this has
something to do with the GPG version?
The downside of this is we will be less likely to notice if GPG
changes again and we start leaking agents like we're in The Matrix
Reloaded. But the real solution to that is containers anyways.
Closes: #233
Approved by: smcv
The API supports this, and it's not hard for us to do in the command
line as well. One possible use case is separating "content
generation" in a separate server.
Related: https://github.com/ostreedev/ostree/pull/223Closes: #227
Approved by: jlebon
When I'm doing local development builds, it's quite common for me not
to want to accumulate history. There are also use cases for this on
build servers as well.
In particular, using this, one could write a build system that didn't
necessarily need to have access to (a copy of) the OSTree repository.
Instead, the build system would determine the last commit ID on the
branch, and pass that to a worker node, then sync the generated
content back.
The API supported generating custom commits that don't necessarily
reference the previous commit on the same branch, let's just expose
this in the command line for convenience.
I plan to also support this rpm-ostree.
Closes: #223
Approved by: jlebon
Just keeping my promise to write more documentation. There could be a
lot more to write here, but I'm trying to get a start done.
Closes: #222
Approved by: jlebon
I'm trying to improve the developer experience on OSTree-managed
systems, and I had an epiphany the other day - there's no reason we
have to be absolutely against mutating the current rootfs live. The
key should be making it easy to rollback/reset to a known good state.
I see this command as useful for two related but distinct workflows:
- `ostree admin unlock` will assume you're doing "development". The
semantics hare are that we mount an overlayfs on `/usr`, but the
overlay data is in `/var/tmp`, and is thus discarded on reboot.
- `ostree admin unlock --hotfix` first clones your current deployment,
then creates an overlayfs over `/usr` persistent
to this deployment. Persistent in that now the initramfs switchroot
tool knows how to mount it as well. In this model, if you want
to discard the hotfix, at the moment you roll back/reboot into
the clone.
Note originally, I tried using `rofiles-fuse` over `/usr` for this,
but then everything immediately explodes because the default (at least
CentOS 7) SELinux policy denies tons of things (including `sshd_t`
access to `fusefs_t`). Sigh.
So the switch to `overlayfs` came after experimentation. It still
seems to have some issues...specifically `unix_chkpwd` is broken,
possibly because it's setuid? Basically I can't ssh in anymore.
But I *can* `rpm -Uvh strace.rpm` which is handy.
NOTE: I haven't tested the hotfix path fully yet, specifically
the initramfs bits.
In some cases (such as `ostree-sysroot-cleanup.c`), the surrounding
code would be substantially cleaner if it was also ported to
fd-relative, but I'm going to do that in a separate patch.
That way these patches are easier to review for mechanical
correctness. I used an Emacs keyboard macro as the poor man's
[Coccinelle](http://coccinelle.lip6.fr/).
Since we hard-depend on GLib 2.40, we can start using GSubprocess.
This is part of dropping our dependency on libgsystem, which is
deprecated in favor of libglnx (as well as migrating things to GLib).
I'd like to encourage people to make OSTree-managed systems more
strictly read-only in multiple places. Ideally everywhere is
read-only normally besides `/var/`, `/tmp/`, and `/run`.
`/boot` is a good example of something to make readonly. Particularly
now that there's work on the `admin unlock` verb, we need to protect
the system better against things like `rpm -Uvh kernel.rpm` because
the RPM-packaged kernel won't understand how to do OSTree right.
In order to make this work of course, we *do* need to remount `/boot`
as writable when we're doing an upgrade that changes the kernel
configuration. So the strategy is to detect whether it's read-only,
and if so, temporarily mount read-write, then remount read-only when
the upgrade is done.
We can generalize this in the future to also do `/etc` (and possibly
`/sysroot/ostree/` although that gets tricky).
One detail: In order to detect "is this path a mountpoint" is
nontrivial - I looked at copying the systemd code, but the right place
is to use `libmount` anyways.
This content currently lives here:
<https://wiki.gnome.org/Projects/OSTree/RelatedProjects>. Moving it
into the manual in Markdown:
- Makes it look better
- It's more useful alongside the rest of the docs
- Is much less crummy in general than the GNOME wiki
GNOME Continuous uses ginstest-runner --report-directory, which causes
the tests to save their tmpdirs persistently. This also means the
result directories didn't match the `/(var/)?tmp` regexp, which broke
the ostree tests in GContinuous.
Fix this by simply asserting that the tmpdir either has `.tmpdir` or
nothing.
We were considering using this for Docker integration, but we may end
up going a different architectural path. Anyways, it doesn't hurt to
have the bindings in here - they can do a few things.
I decided to fork some of the core code from
https://github.com/dradtke/gotk3 because...well, what we really need a
GIR-based core generator but I didn't want to start on the fully
correct thing until we knew we wanted it, and this was a quick hack.
Also, let's make a `contrib/` directory for things like this.
Otherwise $(installed_testdir) is empty so we try to put content in
`/`, which I noticed when trying to build an RPM (it works works fine
`sudo make install`).
%Z only uses seconds, so it's possible that we did the commit
in the same second, which made this test racy.
- Switch to full nanosecond precision using '%.Y' so it always differs
- Fix the inverted `cmp` usage
- Add a missing `ok`
This will allow daemons like rpm-ostree to detect if there are any new
deployments efficiently, in combination with using inotify. If there
are any changes, rpm-ostree wants publish them on DBus.
While we're here, add some changes to start doing unit C testing of
the sysroot API.
I want to be able to easily test the C API on actual data in an OSTree
repo. The shell `libtest.sh` has code to generate it. Bridge the two
worlds by introducing a little `libostreetest` library which has a C
API which spawns a shell that runs things in `libtest.sh`.
Yes, this is about as beautiful as it sounds, which is to say, it's
not. But it works!
Note while we were here, I realized we were actually now creating
*two* tmpdirs per test in `make check` because the tap driver was
already doing that. Unify it so we know the C code can rely on it.
And change the command line to use it. rpm-ostree had a copy
of this code, and thus there's a clear reason to have an API.
While we're moving this into API, ensure the mtime on deploy is bumped
after an osname is created, so that daemons like rpm-ostree can notice
changes. (In reality, creating the directory should do this, but
let's be double sure)
This allows other processes (e.g. rpm-ostreed) to monitor for external
changes (e.g. if someone does `ostree admin undeploy`) in a relatively
sane fashion.
Specifically, I'm trying to fix:
https://github.com/projectatomic/rpm-ostree/issues/220
OSTree's code for testing predates the `glib-tap.mk` making its
way into GLib. Let's switch to it, as it provides a number
of advantages.
By far the biggest advantage is that `make check` can start to run
most of the tests *in addition* to having them work installed.
This commit keeps the installed tests working, but `make check` turns
out to be really broken because...our TAP usage has bitrotted to say
the least. Fix that all up.
Do some hacks so that the tests work uninstalled as well - in
particular, `glib-tap.mk` and the bits encoded into
`g_test_build_filename()` assume *recursive* Automake (blah). Work
around that by creating a symlink when installed to loop back.
I plan to use this in rpm-ostree at least for two reasons:
- To find the mtime on the repo
- To use the tmp/ directory to stage content (but we should eventually
add a better API)
As rpm-ostree evolves, it keeps driving API additions to libostree.
This creates a relatively tight coupling.
However, if delivering via e.g. RPM, unless one manually remembers to
increment the `Requires:` in the spec file, it's possible for the two
to become desynchronized.
RPM handles versioned symbols and will ensure a dependency if the
application starts using a newer version.
To implement this, switch to `-fvisibility=hidden`, along with an
annotation in the header, and finally add a `.sym` file.
This matches what other projects like systemd and libvirt do.
Although rather than attempting to retroactively version symbols, glom
them all onto the current one.
I see when analyzing a delta here that due to byteswapping a negative
compression ratio of 540%, 66%, and 28%. Let's arbitrarily pick 20%
as a threshold for detecting byetswapping.
If the average object size is greater than 4GiB, let's assume we're
dealing with opposite endianness. I'm fairly confident no one is
going to be shipping peta- or exa- byte size ostree deltas, period.
Past the gigabyte scale you really want bittorrent or something.
Some Docker layers are just metadata in the `layer.json`. If one is
mapping Docker layers to OSTree commits, one needs to create a dummy
root directory, because OSTree doesn't support metadata-only commits.
Let's just push that logic down here because it's easier than special
casing it in higher levels.
One of the design goals with deltas was not just wire efficiency,
but also having all the data up front about how much data would
be transferred before starting.
Let's expose that better by adding a `dry-run` option to the pull API.
This requires static deltas to be useful. Basically we simply call
the progress callback once with the data from the superblock.
For a production release repository, most OS vendors would want
to just always use static deltas. Add the ability for the pulls to
require it.
(I think I'll also add a summary key for this actually in addition,
so the repo manager can force it too)
This is a more flexible version of the previous
ostree_repo_write_archive_to_mtree() which took a file reference.
This has an extensible options structure, and in particular
now supports `ignore_unsupported_content`.
I plan to use this for importing Docker images which contain device
nodes. (There's no reason for container images to have those, so
we'll just ignore them).
Also here, just like the export variant, the caller is responsible for
setting up libarchive.
I was going to add new API for importing, and it was really confusing
that what I think of now as import and export both had "write" in the
name. It's just clearer to talk about the direction.
At the same time, include `Export` in the options structure.
This isn't an ABI break as the API isn't in a release.
I was getting a compilation error with the GCC hardening flags which
look for a missing mode with `O_CREAT`. The right fix here is to drop
`O_CREAT`, as truncate() should throw `ENOENT` if the file doesn't
exist.
For rofiles-fuse. Eventually what we should really do is split out
the shared library from the binaries? A minimal system shouldn't need
rofiles-fuse, it's mainly for doing layered packages and that sort of
thing.
I don't know why we didn't do this a long time ago. This extends the
pull API to allow grabbing a specific commit, and will set the branch
to it. There's some support for this in the deploy engine, but there
are a lot of reasons to support it for raw pulls (such as subset
mirroring cases).
In fact I'm thinking we should also have the override-version logic
here too.
NOTE: One thing I debated here is inventing a new syntax on the
command line. Git doesn't seem to have this functionality (probably
because it'd be rarely used). The '@' character at least doesn't
conflict with anything.
Anyways, I wanted this for some other test cases. Without this,
writing tests that go between different commits is more awkward as one
must generate the content in one repo, then pull downstream, then
generate more content, then pull again. But now I can just keep track
of commit IDs and do exactly what I want without synchronizing the
tests.
At the moment I'm looking at using rpm-ostree to manage RPM inputs
which can then be converted into Docker images. It's most convenient
if we can stream directly out of libostree rather than doing a
checkout + tar combination.
There are also backup/debugging etc. reasons to implement `export` as
well.
While it's not strictly tied to OSTree, let's move
https://github.com/cgwalters/rofiles-fuse in here because:
- It's *very* useful in concert with OSTree
- It's tiny
- We can reuse OSTree's test, documentation, etc. infrastructure
One thing to consider also is that at some point we could experiment
with writing a FUSE filesystem for OSTree. This could internalize a
better equivalent of `--link-checkout-speedup`, but on the other hand,
the cost of walking filesystem trees for these types of operations is
really quite small.
But if we did decide to do more FUSE things in OSTree, this is a step
towards that too.
Now we display stats on the individual parts, such as the blob size
and the number of each type of opcode. Most interesting to me is
things like how many bsdiff opcodes there are vs new objects, etc.
We had code to deal with opening/checksumming/decompressing static
deltas in a few places. I'd like to teach `ostree static-delta show`
how to display more information, and this will allow it to just use
`_ostree_static_delta_part_open()` too.
Right now though, almost all of the details of deltas are private, so
we can't do the "honest thing" and have the command line just use the
shared library.
Eventually some of this should appear in the API, but for now add
command line which is useful for debugging.
I'd like to incrementally convert all of `ostree-repo*.c` to
fd-relative usage, so that we can sanely introduce
`ostree_repo_new_at()` which doesn't involve GFile.
This one is medium risk, but passes the test suite.
I don't much like Docbook (and am considering converting the man pages
too), but let's start with the manual.
I looked at various documentation generators (there are a lot), and
I had a few requirements:
- Markdown
- Packaged in Fedora
- Suitable for upload to a static webserver
`mkdocs` seems to fit the bill.
This is preparation for introducing a `mkdocs` manual under `doc/`
which should be significantly more useful for the world at large than
the minimal manual that exists there now.
I noticed in the static deltas tests, there were some tests that
should have been under `-o pipefail` to ensure we properly propagate
errors.
There were a few places where we were referencing undefined variables.
Overall, this is clearly a good idea IMO.
The original intention here was that we'd keey around a copy of the
file so that grub2 could eventually learn how to do atomic updates by
checking for a "fully written" marker in the *new* file, and if it
didn't exist, falling back to grub2.cfg.old.
I haven't yet proposed that upstream, but we might as well stop
deleting the file since it's useful as a backup at least.
Reported-by: Gatis Paeglis
Colin Walters