vika_nezrimaya has quit [Ping timeout: 265 seconds]
{^_^} has quit [Remote host closed the connection]
{^_^} has joined #nixos-aarch64
lopsided98 has joined #nixos-aarch64
<gchristensen>
instead of trying to figure it out, I just deleted some code to make it shorter
<clever>
gchristensen: my local hydra is also failing to build...
<clever>
../libhydra/hydra-config.hh:17:24: error: invalid initialization of reference of type 'const Path&' {aka 'const std::__cxx11::basic_string<char>&'} from expression of type 'std::optional<std::__cxx11::basic_string<char> >'
<clever>
if (pathExists(hydraConfigFile)) {
<clever>
../libhydra/db.hh:15:60: error: no matching function for call to 'getEnv(const char [10], const char [21])'
<clever>
auto s = getEnv("HYDRA_DBI", "dbi:Pg:dbname=hydra;");
<clever>
a `git pull` on hydra appears to have fixed it
<clever>
your initrd thing is also still building
<gchristensen>
making the file a bit smaller seems to have helped
<clever>
nas...> Job for hydra-init.service failed because the control process exited with error code.
<clever>
Mar 09 22:53:46 nas hydra-init[26515]: schema upgrade failed: main::run_(): DBI Exception: DBD::Pg::db do failed: ERROR: column "jobset_id" contains null values at /nix/store/q8j0zzvx49i55m6afxp589f0byv98am5-hydra-0.1.0.0000000000000000000000000000000000000000/bin/.hydra-init-wrapped line 70
<clever>
gchristensen: wasnt that supposed to be fully automatic (with possible extended downtime)?
<gchristensen>
yeah as long as you migrate to an interim commit
<clever>
i just went to head of master, and it failed
<gchristensen>
yah you'll need to pick an interim commit
<clever>
which one would be good?
<gchristensen>
like c4cc72f94449564b56ad9f71d530ee5917af3a97
<gchristensen>
but if you don't have abig db, ust run hydra-migrate
<clever>
hydra-migrate: command not found
<gchristensen>
tab-complete
<gchristensen>
?
<clever>
i did, it wasnt found
<clever>
[hydra@nas:~]$ ls /nix/store/q8j0zzvx49i55m6afxp589f0byv98am5-hydra-0.1.0.0000000000000000000000000000000000000000/bin/
<clever>
Mar 09 23:31:00 nas hydra-queue-runner[31475]: possibly transient failure building ‘/nix/store/2vggq7y55l4587nqg590s4mzwj90gws6-bootdir.drv’ on ‘builder@system76.localnet’: unexpected end-of-file
<clever>
gchristensen: and now hydra is failing all builds
<gchristensen>
heh
<clever>
might be ssh related, let me check...
<clever>
ehhh, some are passing, weird
<clever>
[9084414.842244] starman worker [4180]: segfault at 1054 ip 00007f8c9c8a73b0 sp 00007fff2b819360 error 4 in libnixrust.so[7f8c9c8a5000+1f000]
<clever>
gchristensen: and hydra-server managed to segfault in rust!
<clever>
gchristensen: oh, and ive also heard recently about arm cores, how you can choose to make an arm cpu, that lacks 32bit EL1, but still has 32bit EL0
<clever>
gchristensen: so the kernel must be 64bit, but the userland can be 32 or 64
<clever>
and that can be cheaper (in terms of die area) then a cpu that can be 32 or 64 in every level
<Danct12[m]>
well, afaict it's the first device ever on pmos to have this patch
<samueldr>
I haven't verified those assertions yet
<samueldr>
I think what's happening is that the framebuffer logo doesn't show up as wrong RGB/BGR since G is in the middle
<samueldr>
and that wayland handles that format just fine
<samueldr>
but I've been so far removed from postmarketOS since I added asus-z00t that I can't say for sure if that's right
<samueldr>
it would be weird that the same kernel (asus-z00t)
<samueldr>
wouldn't have the same issue on postmarketOS
<Danct12[m]>
fbsplash seems to handle that just fine too
<samueldr>
it could be
<samueldr>
I was relying on ply-image (originally from plymouth, though forked by chromeos) for splashes
<Danct12[m]>
alright, so im building nixos for santoni
<Danct12[m]>
here's a problem when i build it:
<Danct12[m]>
error: The option `mobile.system.system' is used but not defined.
<samueldr>
you'll need `mobile.system.type = "android";` to be set for the device
<samueldr>
"used but not defined" means that something relies on the value of the configuration, but nothing set its value (and it has no default value)
<samueldr>
uh
* samueldr
is wrong
<samueldr>
I should have read
<Danct12[m]>
mobile.system.type it is set in default.nix, so that's not a issue
<samueldr>
mobile.system.system <- is it possible you doubled `system` somewhere, e.g. mobile.system = { system = "..."; };
<samueldr>
do you have that as a branch I can look at?
<Danct12[m]>
pushing it
<samueldr>
ah, I think I see, `modules/hardware-qualcomm.nix`
<samueldr>
I figure you set soc = "qualcomm-msm8940"; but it's not in that list
<samueldr>
the first part makes the sdm660 option, in that example, the second part configures `system.system` and some quirks when that option is enabled
<Danct12[m]>
oh right, that's the reason
<Danct12[m]>
quirks.qualcomm.msm-fb-handle.enable is imo msm-fb-refresher?
<samueldr>
right
<samueldr>
that commit was from before I dropped fb-handle
<samueldr>
fb-handle didn't work right with LVGL
<samueldr>
(speaking of, it would be much better if I simply added that into LVGL for early boot stuff)
<Danct12[m]>
question, how can i set a kernel config?
<Danct12[m]>
is there anything like kconfig edit like in pmbootstrap?
<samueldr>
sadly not yet, you should be able to drop into a nix-shell to make menuconfig, and configure stuff just right, but I don't have docs for that yet because I haven't figured it out myself yet
<samueldr>
last time I tried it assumed I wanted an x86_64 kernel
<samueldr>
what I've been doing is extract the config from a boot.img's kernel, starting from there enable the options I know are needed, build and hopefully boot all the way
<Danct12[m]>
any list of what to enable and what not to enable?
<samueldr>
I'll be extremelt annoying: not yet :(
<Danct12[m]>
nvm, there isn't any docs yet
<samueldr>
extremely annoying*
<samueldr>
yeah
<samueldr>
though I have an android device I'm keeping untouched to write docs about every little details
<samueldr>
last time I added an android kernel config I used `vimdiff` with a kernel from around the same era
<samueldr>
and one that has the same base kernel version
<Danct12[m]>
that would be nice to see some docs, which would then help more people to get involved in the project
<samueldr>
it should show some things like FRAMEBUFFER FB, USERNS and such as enabled
<samueldr>
you're definitely right, it's on the roadmap
<Danct12[m]>
real linux for phones are kinda underrated, until the pinephone and librem era
<Danct12[m]>
and even then not everyone can afford one of these devices :P
<samueldr>
I made sure to port to a variety of android devices for that one simple goal: use a device you already own
<insep[m]>
fun fact: msm-fb-refresher is not only for msm devices :)
<samueldr>
fun!
<samueldr>
what did you end up needing it with?
<samueldr>
mtk?
<Danct12[m]>
first time having a nixos chroot setup just for mobile-nixos development.. everything is kinda new and weird lol
<Danct12[m]>
really not used to the new development environment
<samueldr>
in theory a chroot shouldn't be needed, only having nix installed, though that does mean the system has /nix and nix installed on it
<samueldr>
and if you use a chroot (or a VM) you could also use another distro with nix, but eh, NixOS is quite good :)
<Danct12[m]>
oh so basically you can still have arch, and /nix in the root directory..
<samueldr>
nix can be used on other distros (and on macOS) without being a nuisance to other package managers
<insep[m]>
<samueldr "what did you end up needing it w"> spreadtrum
<samueldr>
(on macOS you will have some issues with building linux things)
<insep[m]>
but it works on mtk too
<Danct12[m]>
oh right, development on a mac kinda sucks.. i guess
<Danct12[m]>
mainly because most of the stuffs are meant to be used with glibc and musl i guess
<samueldr>
it should be doable as long as you have a nix builder, since nix does not require to build on the platform you evaluate on
<samueldr>
as long as you have a linux nix builder*
<samueldr>
and that linux nix builder could be a VM
<samueldr>
your files, and development would happen on the mac side of things, but end up being built on that other machine
<Danct12[m]>
hmm, so that's like distcc, kinda cool
<Danct12[m]>
and the isolated environment is much more easier to uninstall when you don't need it.. or just want to restart the environm
<samueldr>
well, there's no "environment" :)
<Danct12[m]>
`No space left on device`
<Danct12[m]>
s*it
<samueldr>
because of how nix evaluates, any change ends up changing the hash of that derivation, and all derivations depending on it, meaning that in practice you don't get "bad cache"
<Danct12[m]>
37G 35G 0 100% /
<samueldr>
yeah, that will happen :)
<samueldr>
each successful builds are kept
<Danct12[m]>
is there anyway to change this folder in anyway?
<samueldr>
the /nix folder?
<Danct12[m]>
yeah
<Danct12[m]>
i think it's torturing the rootfs part
<samueldr>
since you're not botting from it or relying on it to boot, it could be a mount or bind mount without issues
<samueldr>
not sure if anything would break with a symlink though
<Danct12[m]>
mhmm
<samueldr>
(some software is sneaky and use readlink hapazardly)
<samueldr>
the "dumb" way to change nix to just use another folder will not work as expected, as /nix is part of the inputs, so the resulting hashes end up changing
<samueldr>
there is another way which I'm looking up, trying to buy time by writing sentences in IRC doesn't help :)
<Danct12[m]>
ah, nix-collect-garbage
<samueldr>
--store local?root=some/path can be used with nix-build, it will completely ignore the /nix root, but will use user namespaces and act as if some/path/nix was /nix
<LinuxHackerman>
or --store /some/path for short ;)
<samueldr>
I assumed that was changing the store path in the "dumb" way previously mentioned
claudiii has joined #nixos-aarch64
<LinuxHackerman>
samueldr: nope, that only works via NIX_STORE_DIR iirc
<samueldr>
LinuxHackerman: good to know, I tried to find actual docs about it (thus my question in #nixos-dev)
<Danct12[m]>
anyway to change the build directory somehow?
<Danct12[m]>
everything currently spats out in /run/user/1000 which then would go out of space and stops compiling
<LinuxHackerman>
TMPDIR
<LinuxHackerman>
of the nix-daemon, if one is running
h0m1 has quit [Quit: WeeChat 2.7.1]
h0m1 has joined #nixos-aarch64
<Danct12[m]>
ok_hand
lovesegfault has quit [Quit: WeeChat 2.7.1]
cidkid has joined #nixos-aarch64
<gchristensen>
thefloweringash: I'm thinking this machine should run 4 builders, and have the host run a nix-daemon, using remote building to the 4 VMs on the host
zupo has joined #nixos-aarch64
<misuzu>
gchristensen: a lot of jobs seem to fail due to oom, maybe there is some issue with kernel? i'd check output of free -h on those VMs just to make sure
<gchristensen>
total used free shared buff/cache available