Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
Accidentally broken in last commit.
|
|
|
|
|
|
packagekitd.service will be stopped on system shutdown and by manually
doing so. Thus the main process will be killed eventually and we have
to manually clean those tmporary libzypp files.
This reverts commit d8dd484ddfcea4750495fa92dcd2f580149432e7.
https://bugzilla.opensuse.org/show_bug.cgi?id=1169739
|
|
Ohh TeX Live, you taunt me.
Fixes https://github.com/hughsie/PackageKit/issues/442
|
|
|
|
|
|
Fixes https://github.com/hughsie/PackageKit/issues/436
|
|
|
|
|
|
The previous commit was meant to workaround the fact that PackageKit
doesn't quit normally on a timed exit. Since commit 8b621738 fixes
that, this is no longer needed.
This reverts commit 807f410e28074c8ad2b99682446bbb460fd54adf.
|
|
This makes sure that the main process won't get SIGTERM on shutdown,
for example timed exit after idling for a while. This fix the problem
that libzypp fails to clean up some temporary files when packagekitd
quits.
|
|
|
|
We can use newer Meson this way, and more importantly also test against
a more recet version of APT.
|
|
This resolves #366
|
|
|
|
|
|
https://bugzilla.suse.com/show_bug.cgi?id=1173562
|
|
(CVE-2020-16121)
These functions revealed existence and content type of files, which
allows a non-root user to check existence and content type of any
file on the system, regardless of permission, as the checks are
performed as root.
A correct fix would move those checks into the client, and pass an
fd to the daemon. Here we just hide which failure it is, which we
would need to do anyway, but don't provide an improved version as
that's out of scope for a security issue and requires changes the
reverse dependencies using those functions.
Bug-Ubuntu: https://bugs.launchpad.net/bugs/1888887
|
|
Debs do not have signatures on their own, so they are always
untrusted.
Bug-Ubuntu: https://bugs.launchpad.net/bugs/1882098
|
|
|
|
https://bugzilla.suse.com/show_bug.cgi?id=1175315
|
|
During the conversion from Autotools to Meson, this part was accidentally
lost, which makes the test backend not work quite right.
This change installs the missing files for the test_spawn part of the
test backend.
|
|
|
|
|
|
|
|
Fixes https://github.com/hughsie/PackageKit/issues/428
|
|
|
|
libzypp creates some temporary files under /var/tmp/, somehow they are
not cleaned automatically when zypp backend unloads. Do the cleanup
work manually here.
https://bugzilla.suse.com/show_bug.cgi?id=1169739
|
|
Instead of joining the orignal arguments with spaces and executing
with g_spawn_command_line_sync use g_spawn_sync with a new argv[]
built from exec + arguments so we respect the original command line
arguments
|
|
openSUSE now ships the DNF backend for PackageKit.
|
|
Prevent SQL injection by using sqlite3 statements for SQL queries
when the query is formatted with user provided input.
Signed-off-by: Juho Hämäläinen <juho.hamalainen@jolla.com>
|
|
In case a user asks pkcon to install an already installed package,
so far we returned with:
Fatal error: The packages are already all installed
(detailed string provided by the backend)
and retval 7.
Taking pkcon's manpage into account, retval 5 is more appropriate:
RETURN VALUES
0
Success
5
Nothing useful was done.
7
The transaction failed, see the detailed error for more information.
|
|
Updating without synchronising databases is like pacman -Syu without the pacman -Sy:
- It's normally not possible
- That leaves the system in an inconsistent state
|
|
Previously, in offline updates Tumbleweed is not identified as
Tumbleweed so that updates often fail.
Fix that by building ResPool before the condition check.
|
|
This is `const gchar*`, so pass NULL.
|
|
|
|
|
|
|
|
This reverts commit 0c84d71509e851db20445c747529bd7d3724f081.
This broke the end session dialog in GNOME Shell.
I thought I had that fixed by changing GNOME Shell to autostart
PackageKit when it is not already running:
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1104
But turns out, there are more problems. The update fails if the user has
used dnf since PackageKit last quit, a scenario I did not test. It also
results in GNOME Software getting confused about what software is
installed and what software is not. The issues are discussed here:
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1227#note_784327
It's still desirable to shut down automatically, but it requires more
planning. And shutting down GNOME Software itself is frankly more
important, since it uses a lot more RAM than packagekitd.
|
|
Historically, the backend has internally determined its setup with
static values. However, we generally want PackageKit to load all
repositories defined in all repository directories that DNF normally
searches, since it is not guaranteed to be in /etc/yum.repos.d and
DNF supports multiple repository configuration paths.
We also need the vars to be loaded so that repository definitions
that rely on more than the built-in vars will work.
This bumps our dependency for libdnf to 0.43.1, as we're using APIs
introduced in this release. This required bumping the CI environment
to Fedora 31.
|
|
Currently, PackageKit builds spew implicit function declaration warnings
for usage of realpath(3), which is gated by a _DEFAULT_SOURCE define.
Since we expect it to be usable, let's add the definition so it is.
|
|
|
|
|
|
|
|
alpm: improve dlcb handling
|
|
|
|
When the target name is blank, the callback is generally for the
transaction in general and not for an unknown item. This handles
all instances where libalpm currently emits a progress callback with a blank
name.
|
|
This prevents updating the system syncdbs without updating packages,
which often leads to undesirable behaviour on most alpm-based distros.
|