Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
See: https://github.com/spack/spack/issues/40334
Also:
* Upstream Spack assumes that 'linux-headers' is
always installed on the system, since it is a
dependency of glibc. As 'openssl' requires it,
make the dependency explicit. There are likely
many more occurrences of this that should be
reported and fixed upstream.
|
|
Commit 49431cc5eda8d4e87633c6f201eed7f15df81abf bumped
Python but did not delete the patch rendered obsolete.
See also: #828
|
|
|
|
|
|
Closes: #1104
|
|
Switching to a file-based (as opposed to patch-based)
configuration will be easier for us to maintain. Note
that '/opt/spack/etc/spack/config.yaml' will override
values in '/etc/spack/config.yaml'.
Also:
* change upstream spack.git provider; fixes spack#8,
which gives users immediate access to our patches
if they wish to set up a separate Spack tree
|
|
|
|
Additional changes:
* add 'linux-headers' to makedepends; required by openssl
during clingo bootstrap
* add .post-install file to inform admins about usage
|
|
* ppc requires explicit -latomic; this patch fixes adelie/spack#6
for both the initial clingo bootstrap and 'spack install clingo'
* fix regression where all targets not supported by golang/rust
could not be bootstrapped; this patch fixes adelie/spack#7
* pmmx still requires the archspec patch; fixes #794
* enable debug output during bootstrap build
* disable buildcache by default (do not attempt to pull binaries)
* fix package paths
|
|
|
|
* Make dependencies explicit
https://spack.readthedocs.io/en/latest/getting_started.html#system-prerequisites
* Remove cached compiler preferences; this causes problems
if e.g. 'ccache' is used ('/usr/local/bin/gcc') when built
but the same path does not exist on the installed system.
* Do not attempt to use the buildcache (search remote server
for pre-built packages). We are the only musl-based distro
currrently packaging Spack, and no such binaries exist yet.
|
|
|
|
|
|
|
|
Reported-by: @djt
|
|
Adds modules for most RTCs supported on arm64. We also now build-in
RTC_EFI, which should grab most of the server-class boards.
|
|
It seems prudent to allow users to fetch files from remote servers in
the Base environment.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The new binutils version allows more options on aarch64, pmmx, and
x86_64. Configure them so that native builders don't have `oldconfig`
output.
Fixes: #1091, #1092, #1094
|
|
Additionally, ensure shared libraries are versioned.
This fixes a regression from the autotools-based era
that was not carried into the CMake era.
|
|
|
|
This updates the outputs to 1.0-beta5.
|
|
* Set cap_sys_time on horizon-qt5 to allow setting the clock.
(adelie/horizon#321)
* Move gfxdetect to graphical runlevel. (adelie/horizon#346)
|
|
* Change install scripts to local.d scripts.
* Add a local.d script to make dhcpcd setuid. (adelie/horizon#361)
* Make the Desktop Entry file a real file instead of a heredoc.
* Remove setuid bit from the shell script; we don't need this any more.
* Pull in our new LXQt theme for a better experience.
* Add XFCE subpackage.
* Set PATH correctly for horizon-run-qt5 so it can find mkfs.vfat and
friends.
* Set the clock with chronyd if possible on boot, before the graphical
environment starts.
* Use sudo for partitionmanger. (adelie/horizon#368)
Ref: #1081
|
|
Defines a base LXQt theme that is functional and appealing.
|
|
|
|
This supplies the fonts we want for all desktop environments.
Ref: adelie/horizon#353
|
|
|
|
This update brings some reliability fixes to i915/i965, nouveau, and
also radeonsi. This is the most recent version of the 'amber' branch
which is the last version with support for these chipsets.
There is a way to co-install mesa-amber and mesa, but we will need to
work on that. For now, we ship this for everyone.
The nouveau patch here is sent upstream and fixes context destroys.
|
|
This fixes all the stupid EFI bugs in Horizon.
Upstream: https://github.com/OpenRC/openrc/issues/238
|
|
Add more network drivers to the ppc64 config. Requested by @djt.
|
|
These eval boards set config options that cause pata-macio to panic on
Power Macintosh systems. Right now, the Power Mac is a significantly
more important target for us. The eval board support was new for 5.15
anyway, which hasn't been shipped to any users yet.
|
|
Moved to legacy/ due to filename conflict. See #1089.
Additional work to be done is create a new package for
zlib-ng [1] so as not to depend on upstream zlib, then
ensure that all tests are being run correctly.
It would be a good idea to add a library suffix '-ng'
to avoid conflicts.
[1]: https://github.com/zlib-ng/zlib-ng
|
|
|
|
Fixes: #1090
|
|
|
|
|
|
|
|
Linking now fails on pmmx even with the kitchen sink:
https://wiki.debian.org/ReduceBuildMemoryOverhead
Change the conditional to include all known 32-bit platforms.
Ref: #1012
Fixes: #1057
|
|
TEXTRELs are due to assembly code for FFTs in ffvpx/libavcodec.
Since this is only used on "Android32 FLAC optimisation", it's likely
not worth it for our use cases. Disable it.
Fixes: #1015
|