makerepo.sh failed on RHEL6, because
- realpath was not available
- git-config did not understand the --local option
Fix these issues.
Signed-off-by: Thomas Haller <thaller@redhat.com>
Also, allow makerepo.sh not to be in the root directory of the
dist-git checkout. Instead, your current working directory
must now be inside the git working copy
Signed-off-by: Thomas Haller <thaller@redhat.com>
This script can be used when backporting patches in dist-git for
Fedora and RHEL. Try `./makerepo.sh -h`
Signed-off-by: Thomas Haller <thaller@redhat.com>
This makes NetworkManager independent of <polkit/polkit.h>
development headers and libpolkit-gobject-1.so library.
Instead communicate directly with polkit using its DBUS
interface.
PolicyKit support is now always compiled in. You can control
polkit authorization with the configuration option
[main]
auth-polkit=yes|no
If the configure option is omitted, a build time default
value is used. This default value can be set with the
configure option --enable-polkit.
This commit adds a new class NMAuthManager that reimplements the
relevant DBUS client parts. It takes source code from the polkit
library.
https://bugzilla.gnome.org/show_bug.cgi?id=734146
Signed-off-by: Thomas Haller <thaller@redhat.com>
The module is used for building man pages by generate-plugin-docs.pl script.
1c2174a libnm-util: generate-plugin-docs.pl script for extracting plugin docs
Add libnm to NetworkManager.spec, and try to update the descriptions
of NetworkManager-devel, NetworkManager-glib, and
NetworkManager-glib-devel to make sense in the libnm world...
Create a new clients/ subdirectory at the top level, and move cli/ and
tui/ into it, as well as nm-online.c (which was previously in test/,
which made no sense).
cli/ was split into two subdirectories, src/ and completion/. While
this does simplify things (given that the completion file and the
binary both need to be named "nmcli"), it bloats the source tree, and
we can work around it by just renaming the completion file at install
time. Then we can combine the two directories into one and just have
it all under clients/cli/.
There are three fixes:
1) the config packages shouldn't require the main NetworkManager
package. We explicitly make NetworkManager-glib not
depend on NM itself to ensure that clients can link to it (and thus
have an RPM dependency on -glib) without pulling in NetworkManager
itself. The same should hold true for the config packages, since
consumers of NetworkManager may want to pull them in, but not
necessarily pull NetworkManager in.
For example, GNOME Shell wants to make use of NetworkManager's
connectivity detection *if NM is available*, and this requires
pulling in NM-config-connectivity-fedora, but that shouldn't
pull in NetworkManager itself.
Similarly, we had a problem with RHEL7 making sure that
NM-config-server was installed by default on Server variants but
not on other variants; removing the dependency from the -config
subpackage was the cleanest way to fix that.
Furthermore, nothing in the config sub-packages actually
requires NetworkManager anyway since they are simply config files.
2) nmtui doesn't care what architecture the running NM is since
it communicates solely over D-Bus. So skip the %{?_isa}, just
like we do for other clients (nm-applet, GNOME Shell, etc)
3) Requiring NetworkManager.%{?_isa} from the -devel package
has problems with multilib (see rh #1112367). This is an issue
if you install a 32-bit development environment and try to build
something that uses NetworkManager. Since the -devel package
requires the main package, you can end up either having *both*
NetworkManager.i686 and NetworkManager.x86_64 installed, or you
end up having NetworkManager.i686 replace NetworkManager.x86_64.
Neither one is good. So remove the %{?_isa} bit.
Before, build_clean.sh always required building all NetworkManager
and doing another `make distcheck` before calling rpmbuild.
That is still a good idea, to ensure that we get a proper build.
For some quick testing however, lets speed this up with a new
--dist argument that only calls `make dist`.
Signed-off-by: Thomas Haller <thaller@redhat.com>
Everytime you call mock again, the fstab file will be reset.
So, we have to write it shortly before creating the image.
Signed-off-by: Thomas Haller <thaller@redhat.com>
Kernel changed default filesystem for roots from ramfs to tmpfs.
See http://lwn.net/Articles/559176/ for more information.
The change caused our initramfs to be mounted with size= parameter that equals
to the actual size of unpacked initramfs. That's why mounted / was full and no
space was available (without manual remounting: mount -o remount,size=100% /).
So we explicitly require usage of ramfs to have all RAM of the virtual machine
available for /.
Signed-off-by: Jiří Klimeš <jklimes@redhat.com>
The user must still be member of the 'mock' group though.
Also, hack something, that the current git repositoy will be reused, so
that we don't have to fetch the entire repositoy from upstream.
Signed-off-by: Thomas Haller <thaller@redhat.com>
