IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Prep for "Rust-as-main", where I want to build libdnf statically.
And this really completes the "library thinout" story because
now we avoid dragging our *private* `libdnf.so` into the caller's
address space, which can cause potential conflicts if they're
also linking the system one. (Which could easily occur with
something like gnome-software)
All we were using libdnf for (indirectly via libsolv) is comparing
version strings but librpm can already do that for us.
A long time ago now we started injecting the package list
into the ostree commit.
We were carrying backwards compatibility code to do a checkout
of an ostree commit and extract it, but if we delete that
and have it return an error, it removes the last bits
that end up calling into our internal librpmostreepriv.la.
Now the internal shared library only links to our private `libdnf.so.2`,
`libglnx.la` statically, and libglib.so and libostree.so.
In the future it'd be nice to even drop the linkage to `libdnf.so.2`
because that drags in a lot of stuff, and I suspect e.g. gnome-software
may hit incompatibilities if it also links in the "system" libdnf.
When g-ir-scanner runs, it uses a stub binary to help introspect the
target rpm-ostree library. That binary needs to link to the library, and
of course, transitively, to our bundled libdnf. Since we run the scanner
uninstalled, we use `LD_LIBRARY_PATH` to point the stub at the libdnf
from the build directory.
When we compile rpm-ostree, we use `-rpath` to point it at our libdnf,
which emits a `DT_RPATH` attribute. However, during dynamic linking at
runtime, it turns out that `LD_LIBRARY_PATH` has *lower* precedence than
`DT_RPATH`. This means that if there is already a libdnf at
`/usr/lib64/rpm-ostree`, it takes precedence.
This subtlety is mostly fine to ignore usually, because the stub doesn't
actually run any rpm-ostree business logic, so it doesn't really matter
that it runs against the "technically wrong" `libdnf.so.2`.
Where it becomes obvious something is off however is if we've just
built a new libdnf, and we have new code which references symbols from
the new libdnf that aren't in the stale libdnf at `$libdir/rpm-ostree`.
And this is exactly what was happening in CI (though it's of course
possible to reproduce this locally as well): we were using a new symbol,
`hy_goal_favor`, and building inside the cosa buildroot image, which
already has rpm-ostree installed, and so `ld.so` chose the bundled
libdnf of the installed rpm-ostree when loading the `g-ir-scanner` stub.
Thus why it failed with:
```
./tmp-introspectzh0n2cga/.libs/lt-RpmOstree-1.0: symbol lookup error: .libs/librpmostree-1.so.1: undefined symbol: hy_goal_favor
```
And of course, the root cause here has nothing to do with the stub in
itself. It's actually trivially easy to see the behaviour difference wrt
`LD_LIBRARY_PATH` on installed binaries. Using a shared object
from `python3-libdnf` which links to libdnf without using rpath, we can
see that `LD_LIBRARY_PATH` has an effect:
```
$ ldd /usr/lib64/python3.7/site-packages/libdnf/_transaction.so | grep libdnf
libdnf.so.2 => /lib64/libdnf.so.2 (0x00007fdde789e000)
$ LD_LIBRARY_PATH=./libdnf-build/libdnf ldd /usr/lib64/python3.7/site-packages/libdnf/_transaction.so | grep libdnf
libdnf.so.2 => ./libdnf-build/libdnf/libdnf.so.2 (0x00007fa615048000)
```
Whereas with rpm-ostree:
```
$ ldd /usr/bin/rpm-ostree | grep libdnf
libdnf.so.2 => /usr/lib64/rpm-ostree/libdnf.so.2 (0x00007f7da5271000)
$ LD_LIBRARY_PATH=./libdnf-build/libdnf ldd /usr/bin/rpm-ostree | grep libdnf
libdnf.so.2 => /usr/lib64/rpm-ostree/libdnf.so.2 (0x00007fc905dbd000)
```
And going further down the rabbit hole, `DT_RPATH` is in fact considered
deprecated for this reason; it makes it harder to run with uninstalled
libraries or whatever reason one has to want to override a library (one
can still use `LD_PRELOAD`, though the semantics are different, and it's
less commonly used than `LD_LIBRARY_PATH`). Instead, we should use
`DT_RUNPATH`, which does have lower precedence than `LD_LIBRARY_PATH`,
matching the usual behaviour.
To make the linker emit `DT_RUNPATH` attributes, we have to use the
`--enable-new-dtags` flag. This also then fixes the g-ir-scanner issue
described above.
Right now there's an issue in Fedora with `g-ir-scanner` picking up
`-fstack-clash-protection` from the `sysconfig` Python module and
passing it to `clang`, which doesn't understand this flag yet.
Just work around this by (1) not even building GIR bindings for our
bundled libdnf since there's no need, and (2) overridding the compiler
used by `g-ir-scanner` so it's always `gcc`.
See: https://github.com/projectatomic/rpm-ostree/pull/1787#issuecomment-473971585Closes: #1787
Approved by: cgwalters
This follows up to https://github.com/projectatomic/rpm-ostree/pull/1576
AKA commit 2e567840ca - we now process
treefile inheritance in Rust code. Previously for elements which
reference external files (`postprocess-script` and `add-files`)
we'd hardcoded things to only look in the first context dir.
Now we open file descriptors in the Rust side for these "externals"
as we're parsing, and load them C side. Hence we'll correctly handle
a `postprocess-script` from an included config.
Other advantages are that the include handling was ugly un-typesafe C code
with no unit tests, now it's memory safe Rust with unit tests.
The downside here is I ended up spelling out the list of fields
again - there's probably a way to unify this via macros but
for now I think this is OK.
Closes: #1574
Approved by: jlebon
For https://pagure.io/atomic-wg/issue/299 we need to make it
more convenient to substitute the architecture in an installation
context. I plan to use this API inside `rpmostreepayload` in Anaconda,
so we can substitute the same value of `${basearch}` we use in treefiles
since https://github.com/projectatomic/rpm-ostree/pull/305
Now, you might wonder - why do we need an API wrapping libdnf? It's because
libdnf is not API stable yet. We're just exposing a tiny subset. In theory we
could use the Python dnf bindings in Anaconda, but things get slightly weird if
rpmostreepayload depends on dnf. Perhaps we'll do that down the road, but for
now this a small API surface to maintain (forever).
This change reworks the internal `varsubst` bits to take a pure `DnfContext`,
since we don't want to spin up a whole `RpmOstreeContext` just to do some
string substitutions.
Closes: #877
Approved by: jlebon
So I was trying to hack on my host's copy of rpm-ostree inside a pet
docker container, but ran into a conflict with libhif since dnf uses
it. I think we basically need to *always* build the bundled path,
rather than what I'm doing with CAHC and FADC where it's built as a
regular RPM.
It's not really sustainable right now for us to have both bundled and
not-bundled build paths - and we need to support co-installation with
dnf.
Another major issue is that we want to version lock with libhif -
right now our CI and both CAHC/FADC track libhif master, but that
means everything breaks if libhif breaks and we don't immediately
port.
git submodules solve all of these problems - the same as we're doing
with libglnx.
libglnx is *designed* for use as a git submodule, where as libhif
needs to support being both bundled and not-bundled. So we end up
with some hacks on our side, but I think it's all not too bad. I've
marked build rules with `# bundled libhif` so we know where to find
them later when libhif is stable.
Closes: #357
Approved by: jlebon
We had `src/lib` having its own little private library; I wanted to
use some of it inside `src/libpriv`, so let's consistently have all
private utility code in `src/libpriv`.
Closes: https://github.com/projectatomic/rpm-ostree/pull/147
On the plus side, we share some code between the library and the
binary now. On the downside, because `librpmostreepriv.la` is a
noinst library, its code text is duplicated between the shared library
and binary, at least until we either:
- Have the binary solely use the public shared library (like ostree does)
- Install `librpmostreepriv.so` to e.g. `/usr/lib64/rpm-ostree/librpmostreepriv.so`
without the headers being public
The `QueryResult` class ended up being too awkward; having NEVRA
strings meant for example that clients would have to parse them. It
would be harder to present something like the current `rpm-ostree
upgrade` package diff output.
Now...I debated quite a while before doing this patch. The thing
that's really awful about creating this library is there are *SO MANY*
layers. rpm-ostree → libhif → hawkey → libsolv → librpm. It's enough
to make one question whether one is actually accomplishing anything or
just contributing to a collective insanity...
Let's pretend for now it's the former.
Closes: https://github.com/projectatomic/rpm-ostree/pull/136
This will help build release engineering and other types of tools;
for example, rather than parsing the output of `db diff`, one
should be able to call an API.
Initially, this adds the generic infrastructure for a public shared
library, with a new function call to do the equivalent of `rpm -qa` on
a particular OSTree commit.
Closes: https://github.com/projectatomic/rpm-ostree/issues/117
Closes: https://github.com/projectatomic/rpm-ostree/pull/124
