So what do we do about these builds failing?

[awebeer256]

I've been following uBlue for about two weeks. Do we just wait for Nvidia and/or Fedora to fix it?

(For the lazy, the Fedora 38 builds are failing on line 40 of build.sh: akmods --force --kernels "${KERNEL_VERSION}" --kmod "${NVIDIA_PACKAGE_NAME}". Partial output:

Checking kmods exist for 6.3.5-200.fc38.x86_64 [  OK  ]
Building and installing nvidia-kmod [FAILED]

)

[castrojo]

We need to either wait for it to be fixed or someone needs to PR a fix into the akmods repo.

[akdev1l]

The first step would be to reproduce the build failure on Fedora Workstation - if that's possible then the issue belongs upstream

[awebeer256]

We need to either wait for it to be fixed or someone needs to PR a fix into the akmods repo.

Do you mean the uBlue akmods repo? Because my fork doesn't use that at all, and it still started failing at the same time.

[castrojo]

Yeah there's a bunch of failures across the board with F38 in akmods, main, and nvidia.

[antuan1996]

Is it possible to disable "v4l2loopback" in ublue-os/akmods repository temporarily to fix the build?

[castrojo]

I'm not familiar with how this repo works but if someone has a fix to disable it to unjam the rest I'll review it!

[awebeer256]

I'll repeat myself: my fork of this repo started failing at the same time as this repo did, and mine doesn't use the akmods repo at all. So that can't be the source of the problem.

(Maybe that wasn't completely clear the first time 🥴)

[castrojo]

Sorry I should have finished my thought (was on mobile), there are multiple rpmfusion related failures right now which are causing all the builds to fail so each of them will need investigation.

…

On Mon, Jun 12, 2023 at 11:26 AM Adam Beer ***@***.***> wrote: I'll repeat myself: my fork of this repo started failing at the same time as this repo did, and mine doesn't use the akmods repo at all. So that can't be the source of the problem. — Reply to this email directly, view it on GitHub <https://github.com/ublue-os/base/issues/180>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAJUT3P7SJD4KVL24SF25QTXK4YKTANCNFSM6AAAAAAZCQRUHE> . You are receiving this because you commented.Message ID: ***@***.***>

[EyeCantCU]

Yup. I spent a lot of time looking into the cause. Just waiting on rpmfusion related things to get worked out

[antuan1996]

In awebeer256/ublue-nvidia build logs:

2023-06-08T03:27:54.1615418Z Installing: akmod-nvidia-3:530.41.03-1.fc38.x86_64 (rpmfusion-nonfree)
2023-06-08T03:27:54.1730098Z Building /usr/src/akmods/nvidia-kmod-530.41.03-1.fc38.src.rpm for kernel 6.3.5-200.fc38.x86_64
2023-06-08T03:27:54.1777532Z ERROR: Not to be used as root; start as user or 'akmodsbuild' instead.

In blue-os/akmods build logs:

2023-06-12T13:04:28.4023326Z Installing: akmod-v4l2loopback-0.12.7^20230503g2c9b670-1.fc38.x86_64 (rpmfusion-free-updates)
2023-06-12T13:04:28.4149248Z Building /usr/src/akmods/v4l2loopback-kmod-0.12.7^20230503g2c9b670-1.fc38.src.rpm for kernel 6.3.6-200.fc38.x86_64
2023-06-12T13:04:28.4201899Z ERROR: Not to be used as root; start as user or 'akmodsbuild' instead.

[awebeer256]

@antuan1996 What's your point? That's been happening all along, in both successful and failed builds.

[castrojo]

(Moving this to discussions since it's org-wide)

Ok so @bigpod98 was able to find the issue with akmods not building, so he's fixed that (Thanks!) The 2nd cause (still ongoing) is that quay.io/fedora-silverblue-desktops isn't consistent!

skopeo inspect docker://quay.io/fedora-ostree-desktops/base:38

returns this kernel version: "ostree.linux": "6.3.7-200.fc38.x86_64", which is what we're supposed to be on.

skopeo inspect docker://quay.io/fedora-ostree-desktops/silverblue:38

returns: "ostree.linux": "6.3.6-200.fc38.x86_64",

So the images that are on the older kernels fail since akmods needs the exact kernel versions to match. So hoping they'll get synced up in the next cron run, I've taken an item to report this.

[awebeer256]

hoping they'll get synced up in the next cron run

Looks like that happened, so we're back in business for now.