In the previous version of this derivation, when the command
?- check_installation.
was issued in the swipl prompt, five errors were encountered. This patch fixes
that, so that the installation is correctly checked.
This enables the usage of the command pack_install to install prolog libraries,
which otherwise fails for the lack of libarchive.
Fixes#30891
* Upgrade `graphite-web`, `carbon` and `whisper` from 1.0.2 -> 1.1.3.
* Replaced the deprecated `pythonPackages.graphite_influxdb` with
`pythonPackages.influxgraph.`
* Renamed `pythonPackages.graphite_web` to `pythonPackages.graphite-web`
to be consistent with the Python package name.
* Replaced the unmaintained `pythonPackages.graphite_pager` with
`pythonPackages.graphitepager`
* Moved all new packages from `python-packages.nix` to
`pkgs/development/python-modules`
Since the switch to using python3Packages in commit
72934aa94e, the plugins no longer build
because they end up with a mix of Python 2 and Python 3 packages.
The reason for this is that the Beets package itself uses callPackage to
reference the plugins, however the overrides are not applied there and
thus the plugins end up getting pythonPackages from the top-level which
is Python 2 and beets with Python 3 dependencies.
Unfortunately this is not the only reason for the builds to fail,
because both plugins did not actually support Python 3.
For the copyartifacts plugin, the fix is rather easy because we only
need to advance to two more recent commits from upstream, which already
contain fixes for Python 3.
The alternatives plugin on the other hand is not maintained anymore, but
there is a fork at https://github.com/wisp3rwind/beets-alternatives
which has a bunch of fixes. In 2e4aded366
I already backported one of these fixes to the version from
https://github.com/geigerzaehler/beets-alternatives, but for Python 3
support it's a bit more complicated than just one little fix.
So instead of adding another series of patches which replicate the code
base of the fork and become a maintenance burden, I opted to directly
switch to the fork and remove the patch on our side.
Signed-off-by: aszlig <aszlig@nix.build>
Cc: @domenkozar, @pjones, @Profpatsch
`ocserv` is a VPN server which follows the openconnect protocol
(https://github.com/openconnect/protocol). The packaging is slightly
inspired by the AUR version
(https://aur.archlinux.org/packages/ocserv/).
This patch initializes the package written in C, the man pages and a
module for a simple systemd unit to run the VPN server. The package
supports the following authentication methods for the server:
* `plain` (mostly username/password)
* `pam`
The third method (`radius`) is currently not supported since `nixpkgs`
misses a packaged client.
The module can be used like this:
``` nix
{
services.ocserv = {
enable = true;
config = ''
...
'';
};
}
```
The option `services.ocserv.config` is required on purpose to
ensure that nobody just enables the service and experiences unexpected
side-effects on the system. For a full reference, please refer to the
man pages, the online docs or the example value.
The docs recommend to simply use `nobody` as user, so no extra user has
been added to the internal user list. Instead a configuration like
this can be used:
```
run-as-user = nobody
run-as-group = nogroup
```
/cc @tenten8401
Fixes#42594
TryExec needs absolute path too, otherwise the desktop file will be ignored
unless gnome-session is in PATH, in which case, we would not need to patch
Exec.
GPaste GNOME Shell extension uses GPaste library generated via introspection. Previously, we added the gpaste package to services.xserver.desktopManager.gnome3.sessionPath option, which
added its typelib directory to GI_TYPELIB_PATH environment variable globally, in order for GNOME Shell to be able to find it. This is not very Nix-y, though, so we have decided to patch the code to
append the path to the GI repository search path.
Additionally, the code relies on GPaste’s GSettings schemas, so we had to hard-code the paths to them as well. We ignored the GNOME Shell’s schemas, since they will already be available for the
extension inside GNOME Shell program.
Until now it's impossible to override the attrs of the actual build
instruction for the `termite` package like this:
```
termite.overrideAttrs (_: {
# ...
})
```
This issue occurs since the `termite/default.nix` expressions returns
the `symlinkJoin` expression when I override termite (e.g. to provide a
config file).
I recently patched termite and wanted to apply this patch to my local
termite installation in my system config which is impossible this, so
splitting the wrapper and the build instruction into their own files
makes this way easier to maintian.
When creating a new mobile broadband connection
with the plasma network manager connection editor,
it tries to find a file containing provider
information somewhere in /usr/share/... .
The build recipe contains a patch to fix the lookup path
such that it finds the file in the corresponding package,
probably added due to
https://github.com/NixOS/nixpkgs/issues/9389 .
The actual lookup path is injected into
the patch file with substituteAll.
With commit a31d98f312 ,
the variable name used in subsituteAll changed from
mobile_broadband_provider_info to mobile-broadband-provider-info
(underscores in package names turned into dashes).
Apparently, substituteAll can't handle dashes in variable names.
Consequently, the variable name was no longer resolved.
plasma-nm failed to create new mobile broadband connections;
the connection creator silently exited and logged the error
> plasma-nm: Error opening providers file "@mobile-broadband-provider-info@/share/mobile-broadband-provider-info/serviceproviders.xml"
This commit keeps the dashes in package names, but it
restores the underscores in the variable used by substituteAll,
thereby ensuring the variable gets resolved properly.
atom-beta: 1.29.0-beta1 -> 1.30.0-beta1
The beta version jump seems like an irregular update, but 1.30.0-beta0
was retired rougly 3 hours after release due to an macOS error:
> Fixed an error about a missing string_decoder module when the Git pane
> is displayed on macOS.
This shouldn't affect the Linux x86_64 version, but latest versions are
always nice.
Also refactored the build script, it now takes the version number as
an extra (optional) argument and creates the version number from that.
Also removed `rec` from the main set because it was no longer needed.
