qemu-utils was pulling qemu which is a 900MB dependency. By removing
reference to it (unneeded), we're saving space on our deployments.
qemu-utils is a dependency of cloud-utils
Some history:
The linux kernel v5.18-rc contains the commit
> commit 6e8ec2552c7d13991148e551e3325a624d73fac6
> Author: Jason A. Donenfeld <Jason@zx2c4.com>
> Date: 2022-01-16 14:23:10 +0100
>
> random: use computational hash for entropy extraction
> [...]
which modifies the kernels random number generator.
This change broke VirtualBox 6.1.34 in several ways:
It causes random crashes and filesystem corruption in the guest
(at least on some host CPU models).
More details can be found in the
corresponding ticket in the VirtualBox bug tracker:
https://www.virtualbox.org/ticket/20914
That ticket also contains a patch "vbox-linux-5.18.patch"
for VirtualBox that fixes the problem,
at least for kernels 5.18 and (hopefully) above.
This patch got added to nixpkgs' VirtualBox build recipe with
9c8132494f .
Meanwhile, the kernel patch got backported to LTS kernels.
As the VirtualBox patch contains several `#if RTLNX_VER_MIN(5,18,0)`
clauses to apply the fix,
it can't heal VirtualBox on LTS kernel versions.
The result is that VirtualBox is still broken if used
with linux kernels 5.10 and 5.15
(currenly the default kernel in nixpkgs).
Luckily, VirtualBox developers updated the patch
(now named "vbox-linux-5.19.patch") to not only fix the
problem for the upcoming 5.19 kernel, but also address
backport releases.
The commit at hand replaces "vbox-linux-5.18.patch"
with the new "vbox-linux-5.19.patch",
fixing VirtualBox for LTS kernel releases.
Some source files have "linux-dmabuf-unstable-v1-server.h" header
included, but do not really need it. Thus, that sources do not
have a correctly configured dependency on that header, which is
dynamically generated during the build.
The fix for that error is to remove unneeded #include.
Signed-off-by: Ivan Nikolaenko <ivan.nikolaenko@unikie.com>
* colima: use updated Makefile in build and install phases
* colima: use lima-unwrapped
* colima: fix dynamic version
* colima: 0.4.2 -> 0.4.3
* colima: delete .git folder in postfetch when `leaveDotGit = true`
Otherwise it may lead to non-deterministic behaviour.
Co-authored-by: j-k <dev@j-k.io>
* colima: do not override default buildGoModule phases
The colima Makefile does more or less the same as the `buildGoModule`
implementation. Instead of overriding it and using the Makefile
directly, we reproduce the behaviour desired behaviour by customising
env variables, e.g. `ldflags` and `subPackages`.
Co-authored-by: j-k <dev@j-k.io>
* colima: use `buildGoModule = buildGo118Module;`
Co-authored-by: j-k <dev@j-k.io>
* colima: update meta.description
Co-authored-by: j-k <dev@j-k.io>
* colima: set `CGO_ENABLED = 1` to make VPN connections work
See https://github.com/abiosoft/colima/issues/358
* colima: 0.4.3 -> 0.4.4
* colima: define all ldflags in preConfigure
Co-authored-by: Atemu <Atemu@users.noreply.github.com>
* colima: add version test
* colima: add meta.mainProgram and meta.platforms
* colima: remove unused runCommand input
Co-authored-by: j-k <dev@j-k.io>
* colima: cleanup meta
Co-authored-by: Sandro <sandro.jaeckel@gmail.com>
* colima: drop meta.platforms
because we use the default value of `platforms.darwin ++
platforms.linux`
Co-authored-by: Sandro <sandro.jaeckel@gmail.com>
Co-authored-by: Atemu <Atemu@users.noreply.github.com>
* Update pkgs/applications/virtualization/colima/default.nix
Co-authored-by: j-k <dev@j-k.io>
Co-authored-by: Atemu <Atemu@users.noreply.github.com>
Co-authored-by: Sandro <sandro.jaeckel@gmail.com>
[ 86%] Building CXX object CMakeFiles/xmrstak_opencl_backend.dir/xmrstak/backend/amd/minethd.cpp.o
nvcc fatal : Unsupported gpu architecture 'compute_30'
CMake Error at xmrstak_cuda_backend_generated_cuda_extra.cu.o.Release.cmake:220 (message):
Error generating
/build/source/build/CMakeFiles/xmrstak_cuda_backend.dir/xmrstak/backend/nvidia/nvcc_code/./xmrstak_cuda_backend_generated_cuda_extra.cu.o
nvcc fatal : Unsupported gpu architecture 'compute_30'
make[2]: *** [CMakeFiles/xmrstak_cuda_backend.dir/build.make:84: CMakeFiles/xmrstak_cuda_backend.dir/xmrstak/backend/nvidia/nvcc_code/xmrstak_cuda_backend_generated_cuda_extra.cu.o] Error 1
CUDA support is broken for a while. Let's drop it along with gcc6 requirement.
Without the change archimedes fails to build with latest toolchains as:
ld: archimedes.c:(.text+0xdac7): undefined reference to `rnd'
ld: archimedes.c:(.text+0xdeab): undefined reference to `rnd'
ld: archimedes.c:(.text+0xdf13): undefined reference to `rnd'
ld: archimedes.c:(.text+0xe3ef): undefined reference to `rnd'
ld: archimedes.o:archimedes.c:(.text+0xe456): more undefined references to `rnd' follow
ld: archimedes.o: in function `EMC':
archimedes.c:(.text+0xf11b): undefined reference to `creation'
ld: archimedes.c:(.text+0xf5f5): undefined reference to `creation'
ld: archimedes.c:(.text+0xf826): undefined reference to `creation'
ld: archimedes.c:(.text+0xf9ce): undefined reference to `creation'
The change happens to fix build against clang as well.
