Reverts commit 70efe55031 for kernels that
started failing on bpo. I suspect it's caused by a change in binutils,
since I tried building linux-oneplus-billie2 with gcc6 and got the same
error (after trying other workarounds like disabling selinux for it, but
then it failed elsewhere).
We previously installed dtbs outside of the boot partition, so it
was necessary to copy them from the root partition to boot if you
wanted to utilise them in U-Boot scripts. This is no longer
necessary as we now install dtbs to the boot partition directly, so
remove this workaround and edit the U-Boot script to reflect the
change.
This is possible now that all SDL (1) applications use sdl12-compat
instead of "classic" SDL 1.2. Note that this only is applied to user
interfaces that use Wayland and are launched through tinydm.
See https://gitlab.alpinelinux.org/alpine/aports/-/issues/12739
This device package represents several devices, so use plural form in
pkgdesc of APKBUILD:
pkgdesc="Google Hana Chromebooks"
However, when deviceinfo is already installed, it is intended to represent
only current device, so use singular form here:
deviceinfo_name="Google Hana Chromebook"
Don't build with !tracedeps, as otherwise the isl dependency is not
properly added to the built gccs. In practice this probably doesn't
matter since we only use these old gccs to build kernels in the native
chroots (not via crossdirect in foreign arch chroots), and we also
install a proper gcc in the native chroot that pulls in the isl
dependency. But fix it for correctness.
Related: pmaports issue 1732
[ci:skip-build] Already built successfuly on CI in MR
- update to use librem5-base v52
- install usbguard config / new subpackage
- move xorg.conf to new x11 subpackage
- refactor udev rule install to just install everything available from
librem5-base upstream
- use purism's shipmode script, see:
https://source.puri.sm/Librem5/linux/-/merge_requests/333#note_195511
We already ship Angelfish and shipping 2 browser takes up a lot of
space. Besides Angelfish feels more native in Plasma Mobile (being baed
on Qt and made for mobile) and has a built-in adblocker nowadays so it
seems like the better option.
urfkill hasn't been used for quite a while, no need to keep it around
Motivation for this change is that the image builds of the UI are
currently running out of space so we need to decrease the image size
Actually package pulseaudio is needed instead of pulseaudio-utils.
Package pulseaudio-utils then gets drawn in by postmarketos-base-ui
sub-package pulseaudio.
For Xfce4 this is a cosmetic change because pulseaudio gets installed
as a dependency of pavucontol already. Keeping pulseaudio in the
_pmb_reccomends list makes sense to be clear which pieces are needed.
For MATE an explicit installation of pulseaudio is needed because it
doesn't use pavucontrol.
[ci:skip-vercheck]
[ci:skip-build]: already built successfully in CI
Linux 5.18 brought better support for some x86-based Android tablets,
thanks to work from Hans de Goede. This includes asus-me176c: the
modified ACPI table and the Linux fork is no longer necessary,
it can just run linux-edge from Alpine.
Drop the old Linux 5.4 fork and related packages. Instead of using
the proprietary sound firmware from linux-firmware-intel, switch to
using the open-source(!) "Sound Open Firmware" (SOF).
NOTE: The mainline battery driver seems quite a bit less accurate
than the ugly old driver ported from the downstream driver. Also,
Bluetooth seems to fail on some boots now. Overall it works quite
well though. :)
Since c49d84f4ab1856156d26d59f5c88387776990d24 in Alpine's aports,
the sm.puri.OSK0.desktop service no longer tries to execute
/usr/bin/osk-wayland and instead just executes /usr/bin/squeekboard
directly, thus avoiding this needless indirection. virtboard hasn't
existed in postmarketOS for years at this point and squeekboard is
a hard dependency of phosh, so it's always going to be present
(thus making this check unnecessary) and the fallback if it isn't
wouldn't work anyway.
[ci:skip-build]: already built successfully in CI
