IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Four checkboxes with "Elbrus 801-PC" prefix look a bit too much,
let's put these under a parent one; the counter-argument is that
selecting that parent checkbox is going to provide message telling
that there are conflicts within -- the proper widget would be a
dropbox with the number of seats but we don't do that just yet
in alterator-pkg.
The autologin group goes there too; still a hack that should be
replaced by proper alterator-users multiseat support.
SYSLINUX happens where GRUB happens too, just in case.
The problem at hand is that some branding-*-bootloader tend to
*create* /etc/sysconfig/grub2 without bothering to check that it
existed, thus confusing the subsequent scripts (e.g. 65-serial.sh
tries to run update-grub that's just absent then).
Otherwise it might e.g. add branding-*-bootloader
for no good reason, poisoning $destdir with an
existing and non-empty /etc/sysconfig/grub2
regardless of the actual grub availability
on the platform.
The point is that it's common between the rest of the arches,
no need to keep making my old mistake of growing configuration
trees based on very different "roots" (.e2k-installer this time
which was good for early experiments but not for production).
Also get rid of long-unused distro/alt-workstation-?01 targets,
unified kernel made those obsolete.
And last but not least: *drop* that setup-libgl-dri3-disable,
makes Radeon RX lag (half-recent AMD GPUs seem to lack 2D part
rather altogether, it's emulated through 3D).
There's no need for bootloader on Elbrus platform
since firmware boots Linux just fine; note that
the whole feature is betted omitted since its
script checks for the case when the feature
is requested but not configured properly.
There's no LibreOffice-still on e2k (no 6.x for now,
just a single production 5.x build); and there's no
office (neither LO nor abiword/gnumeric) at all on
riscv64 as of today. These nuances shouldn't bother
release managers each time but should lurk under
this feature's cover.
The culprit is that e2k has exactly one kernel flavour
available at the moment, elbrus-def (there used to be
per-CPU flavours in p8 times -- elbrus-{4c,8c,1cp} --
but these aren't needed anymore), but there are no
std-def/un-def kernels or provides available.
There's more to it, but lacking a kernel is critical.
Maybe other arches might do similar things when needed,
or maybe this gets reverted in favour of a more generic
approach; but patching all starterkits or messing with
p9.mk per-arch (while leaving out regular builds or
any other image that might just build) looks worse.
Wonder if installer-feature-quota-stage2 was a problem
on other arches but having it in LIVE_PACKAGES fails
image build for me:
livecd-installer-features: Conflicts: installer-common-stage2
and having it INSTALL2_PACKAGES results in another startup
problem (seems like quotas try to start a bit too early);
not that this one is a blocker but spoils the party nevertheless.
NB: relevant package versions are the same.
Also avoid workstation/extras pkglist: it's strongly hardwired,
seems like last-minute implementation :-/
These have been added to the image description
that was x86-specific initially -- we're lucky
that ARM-related bits are quite minor here but
let's keep the mixin away for the sake of e2k
for now: it's built in a different enough way
at the moment.
xorg-drv-mga2 is MCST Graphics Array driver for 2D part
of videochip found in Elbrus-1C+ APU (also to be a part
of Elbrus-2C3).
xorg-drv-smi is another fork of xorg-drv-siliconmotion
that's absent in p9/sisyphus; *might* be of some interest
to mipsel folks (they've got yet another fork it seems).
Ah, and do not override targets on e2k.
Make sure acpid will be running as acpid-events-e2k relies on it
(and power button handling relies on this one).
NB: need to do that before 10-services.
This commit is packing a bunch of changes that went
into RELENG-e2k-workstation during its forked
development; it covers three distinct things:
- boot.conf preparation for firmware to load the image;
- fixup of packages common for e2k images;
- major cleanup of obsolete/unused cruft
(unified kernel and reliable xorg autodetection
allow us to move away from canned machine-specific
configurations for the most part).
installer-feature-setup-plymouth needed make-initrd-plymouth.
make-initrd-plymouth missing for not X86 Arches in alt-server.
This resulted in a make-initrd error when installing on non-X86 Arches.
