Many packages have some kind of flag indicating whether or not to build with
systemd support. Most of these default to `stdenv.isLinux`, but systemd does
not build on (and is marked `broken` for) `isStatic`. Only a few packages have
the needed `&& !isStatic` in the default value for their parameter.
This commit moves the logic for the default value of these flags into
`systemd.meta.{platforms,badPlatforms}` and evaluates those conditions using
`lib.meta.availableOn`.
This provides three benefits:
1. The default values are set correctly (i.e. including `&& isStatic`)
2. The default values are set consistently
3. The way is paved for any future non-Linux systemd platforms (FreeBSD is
reported to have experimental systemd support)
checkInputs used to be added to nativeBuildInputs. Now we have
nativeCheckInputs to do that instead. Doing this treewide change allows
to keep hashes identical to before the introduction of
nativeCheckInputs.
Add a new U-boot flavor for LibreComputer's AML-S905X-CC (Le Potato).
Board's SoC is based on Trusted Firmware-A architecture and needs extra
closed-source early bootloaders to be combined with U-boot in order to
boot properly, similar to Odroid-C2 board.
Proprietary blobs and related tooling are fetched from
github.com/LibreELEC/amlogic-boot-fip and used in postBuild along with
build output.
Hardware: https://libre.computer/products/aml-s905x-cc/
U-boot doc: https://u-boot.readthedocs.io/en/latest/board/amlogic/libretech-cc.html
Note that cups-pdf refuses to run without root privileges.
To use the binary, one has to either convince cups to
call it with root privileges, or install it suid root.
Also note that currently, this cups-pdf-fork produces
small pdfs with selectable text, as promised.
However, copying the text produces "garbled" text
(characters are randomly reassigned).
This is a known issue and I don't know how to fix it:
https://github.com/alexivkin/CUPS-PDF-to-PDF/issues/7
* Rename `cups-kyodialog3` -> `cups-kyodialog`
* Update from 8.1601 (2016) to 9.2 (2022) with support for more and
recent printer models (170 vs. 373 PPD files)
* Make Qt GUI optional to reduce closure size, Qt4 -> Qt5
* Avoid full paths to kyodialog filters in PPD files. A PPD file is
copied to /etc/cups/ppd/ for each configured printer and is not
updated when the cups-kyodialog package is updated, thus,
potentially creating dangling references to /nix/store or invoking
the old version of the cups filters.
The cups.socket unit shouldn't be part of cups.service: stopping the
service would stop the socket and break subsequent socket activations.
See https://github.com/apple/cups/issues/6005
`foomatic-db-ppds` uses the Foomatic database from the packages
`foomatic-db` and -- optionally -- `foomatic-db-nonfree`
and the perl modules from `foomatic-db-engine`
to generate about 8,800 ppd files.
The general structure of the build recipe is as follows:
* Merge `foomatic-db` and `foomatic-db-nonfree` into
one package that represents the Foomatic database.
The package `foomatic-db-nonfree` is optional
as it taints the result license-wise;
it will only be used if `withNonfreeDb`
is to to `true` in the `callPackage` call.
We create a tiny setup hook script that provides the combined
database and sets an environment variable pointing to the
database direcotry, which is expected by the foomatic engine.
* The final package's license and version are computed
from the licenses and versions of the database packages.
The license is set to `free` if each database-providing
package has a free license, and to `unfree` otherwise.
The version is simply the highest version
of the database-providing packages.
* The final package uses `foomatic-compiledb`
from the `foomatic-db-engine` package to extract
all ppd files from the database packages.
`patchPpdFilesHook` is used to patch most
executable invocations in the ppd files
so that they point to nix store paths.
Finally, ppd files are gzipped to reduce storage
(from about 550 MiB to 90 MiB installed).
The "nonfree" version of the package, i.e. the version that is
based on `foomatic-db-nonfree` in addition to `foomatic-db`,
contains about 120 additional ppd files
compared to the "free" version.
Since the "free" version already produces about 8,700
ppd files and hydra won't build the "nonfree" version,
the commit adds two package variables to `all-packages.nix`:
* `foomatic-db-ppds` is based on `foomatic-db` only
* `foomatic-db-ppds-withNonfreeDb`
is also based on `foomaitc-db-nonfree`
The package introduced by this commit
is the result of combining other packages;
it is not the build product of a simple source tarball.
While it would also be possible to perform the ppd file
generation directly in the build process of the database
packages, this would yield further complexity as the
`foomatic-db-nonfree` package needs to be combined with the
`foomatic-db` package before ppd file extraction is possible.
There is no upstream product with a name that
could/should be used for the `name` attribute,
the variable name, or for the filename in nixpkgs.
Similar packages have different names across distributions:
* https://repology.org/projects/?search=openprinting
* https://repology.org/projects/?search=foomatic
The name `foomatic-db-ppds` seems to be most common
(albeit not really *that* common):
* https://repology.org/project/foomatic-db-ppds/versions
At least openSUSE splits their corresponding
package into multiple "binary" packages
(similar to our multi-output packages):
* https://build.opensuse.org/package/binaries/Printing/OpenPrintingPPDs/openSUSE_Tumbleweed
I considered something similar.
However, after doing some statistics,
I concluded that it's not worth the effort:
The biggest dependencies (`perl` and `cups-filters`) are
already present on most NixOS systems, and they cannot
be "split away" easily since it cannot be
done along a canonical line (e.g. printer driver).
Splitting directly by dependency risks that ppd files
unexpectedly "move from output to output" on package updates;
disappearing ppd files can be quite annoying for package users.
`foomatic-db-nonfree` contains -- similar to its sister
package `foomatic-db` -- knowledge about printers,
drivers, and driver options from OpenPrinting in xml files.
It needs to be combined with the `foomatic-db`
package to yield a working database.
It also provides about 100 (gzipped) ppd files.
In contrast to `foomatic-db`, this package contains
data that is provided under non-free licenses.
Quoting
https://github.com/OpenPrinting/foomatic-db-nonfree/blob/master/README
> This is a repository of PPD and Foomatic XML files that may
> have restrictions that keep them from being used on a variety
> of machines for licensing and other non-technical reasons.
ppd files can be found in
`$out/share/cups/model/foomatic-db-nonfree`.
The subdirectory `foomatic-db-nofree` is used to avoid conflicts
with other packages that might be combined with the package
`foomatic-db-nonfree` in `config.services.printing.drivers`.
ppd files in this package are *not* patched
to call executables from the nix store:
The only executable "rastertophaser6100" that is
called from ppd files isn't available in nixpkgs.
There is a daily snapshot of a source archive available at
https://www.openprinting.org/download/foomatic/ .
However, these files rotate daily and
cannot be used as a stable download source.
So we rely on OpenPrinting's Github repository
instead and pinpoint a fresh commit.
Note that the current version is from 2015,
so updates are unlikely.
`foomatic-db` contains the collected knowledge about printers,
drivers, and driver options from OpenPrinting in xml files.
It also provides thousands of ppd files.
The build process patches those files to reference executables
in the nix store and gzips them to reduce storage
(from about 670 MiB to 90 MiB installed).
In contrast to the sister package `foomatic-db-nonfree`
(packaged in a follow-up commit),
this package only includes files published under a free license.
Most files are published under the GPL,
some under the MIT license.
For details see
https://github.com/OpenPrinting/foomatic-db/blob/master/COPYING .
To encompass the different licenses, the package's
license is set to "free" as the common denominator.
ppd files can be found in `$out/share/cups/model/foomatic-db`.
The subdirectory `foomatic-db` is used to avoid conflicts
with other packages that might be combined with the package
`foomatic-db` in `config.services.printing.drivers`.
There is a daily snapshot of a source archive available at
https://www.openprinting.org/download/foomatic/ .
However, these files rotate daily and
cannot be used as a stable download source.
So we rely on OpenPrinting's Github repository
instead and pinpoint a fresh commit.
`foomatic-db-engine` contains several perl scripts to parse
and process XML files from the Foomatic database packages.
It can be used to extract ppd files,
which will be accomplished in a follow-up commit.
The package also contains scripts
to handle print queues and jobs.
It can -- optionally -- talk to the local cups server,
to network printers and to SMB print servers.
The build recipe contains switches to enable these features;
however, they are not needed when generating ppd files.
There is a daily snapshot of a source archive available at
https://www.openprinting.org/download/foomatic/ .
However, these files rotate daily and
cannot be used as a stable download source.
So we rely on OpenPrinting's Github repository
instead and pinpoint a fresh commit.
ppd files in the `samsung-unified-linux-driver`
package invoke these filter commands:
* pstosecps
* pstospl
* rastertospl
* pstosplc
The paths to all of those commands excluding the last one got
patched with their absolute paths during the build process.
This commit adds the last one to
the list of commands to be patched.
The `sed` script in the `installPhase` is removed.
Instead, the setup hook `patchPpdFilesHook` is
used to patch the path to the filter executables.
The result should essentially be the same.
Comparing the generated ppd files showed no difference,
short of the package's hash in the absolute paths and
added newline characters at the end of some ppd files.
Missing newline characters at the end of the last line are
apparently added by `awk`; this shouldn't affect functionality.
The new package also contains a `propagated-build-inputs`
file which propagates the package itself.
This ensures the package is available whenever
a ppd file is singled out by another package.
The `sed` script in the `installPhase` is removed.
Instead, the setup hook `patchPpdFilesHook`
is used to patch the path to `rastertokpsl`.
The result should essentially be the same.
Comparing the generated ppd files showed no difference
(short of the package's hash in the absolute paths).
The new package also contains a `propagated-build-inputs`
file which propagates the package itself.
This ensures the package is available whenever
a ppd file is singled out by another package.
The `cups-drv-rastertosag-gdi` CUPS printer driver
package provides a filter executable `rastertosag-gdi`
and two ppd files that reference the filter file.
The commit at hand uses `patchPpdFilesHook` to
replace calls to the filter executable with
absolute paths to increase package purity.
To consistently return the latest version number even when it is no longer
at the bottom of the alphanumeric list on the Jitsi stable download page.
Based on fix by @sbruder in 13760b87d2
*Flags implies a list
slightly relevant:
> stdenv: start deprecating non-list configureFlags https://github.com/NixOS/nixpkgs/pull/173172
the makeInstalledTests function in `nixos/tests/installed-tests/default.nix` isn't available outside of nixpkgs so
it's not a breaking change
Fixes:
```
.../aarch64-unknown-linux-gnu-ld.bfd: warning: /build/source/build/rk3399/release/bl31/bl31.elf has a LOAD segment with RWX permissions
.../aarch64-unknown-linux-gnu-ld.bfd: warning: /build/source/build/rk3399/release/bl31/bl31.elf has a LOAD segment with RWX permissions
.../aarch64-unknown-linux-gnu-ld.bfd: warning: /build/source/build/rk3399/release/bl31/bl31.elf has a LOAD segment with RWX permissions
make: *** [Makefile:1306: /build/source/build/rk3399/release/bl31/bl31.elf] Error 1
```
See: https://developer.trustedfirmware.org/T996
This commit fixes the problem reported in issue #177572.
Turns out that re-adding this line, which was removed in commit
526238b288, fixes the problem.
The CUPS backend was not built since gutenprint requires libusb1, not 0.1 (at least since ~2014).
With this change CUPS detects printers relying on the pure gutenprint backend (like some dye sublimation printers).
The arm-trusted-firmware/default.nix expression exposes
`buildArmTrustedFirmware` and its `version?"2.7"` field to
`top-level/all-packages.nix`. Unfortunately it doesn't work.
Changing the version field doesn't change what version of the ATF
source code is used. Attempting to "lock" an installation to a
specific version by overriding this field (e.g. version="2.7") won't
work either; when nixpkgs bumps the version to 2.8 the user will end
up building the 2.8 source code but the resulting expression will be
labeled misleadingly in the store:
```
/nix/store/eeee...-arm-trusted-firmware-2.7/
```
**using the 2.8 source code**. So not only does `version` not lock
the version, it will actually *conceal* the fact that the underlying
source code has been upgraded!
Let's just remove the `version` field. It doesn't work and never did.
https://github.com/NixOS/nixpkgs/pull/185004#discussion_r939526830
Co-authored-by: Sandro <sandro.jaeckel@gmail.com>
