CI kills the build of Linux kernel (bug: kernel compile output silent without V=1)

mentioned in merge request !3778 (merged)

By Petr Hodina on 2023-03-04T16:55:17

make the build output the information into the console (files being compiled, etc.) - go through the changes applied on top of mainline

I don't think this can be a regression in the Linux kernel sources. It started happening with the Exynos4 kernel too despite zero changes in the kernel sources.

By Newbyte on 2023-03-04T22:46:28

changed title from Pinenote: CI kills the build of Linux kernel to CI kills the build of Linux kernel

By Oliver Smith on 2023-03-05T15:20:20

Removed "Pinenote" from the title, as this affects multiple kernels.

So far I've seen from recent builds:

Fails without V=1:

6.1.7	linux-postmarketos-exynos4
5.18.5	linux-google-nyan
5.18.0	linux-postmarketos-mediatek-mt8183
5.17.0	linux-pine64-pinenote

Builds fine without V=1:

6.2.0	linux-postmarketos-qcom-sm7150
4.19.95 linux-oneplus-billie2

Looking at blame and history for the top Makefile, there are several commits related to silent mode:

https://github.com/torvalds/linux/commit/6ae4b9868a8f723cae2600722eea033fafadd399
https://github.com/torvalds/linux/commit/8962b6b475bddc011c414f40ffd02f0ed4e02771
https://github.com/torvalds/linux/commit/fc5d57a9a4fe3e3662c4ec4e84772dcea64979ca
https://github.com/torvalds/linux/commit/4bf73588165ba7d32131a043775557a54b6e1db5
- between 6.1-rc8 and 6.1 proper
- "kbuild: Port silent mode detection to future gnu make."
- This one is particularly interesting, apparently make 4.4 breaks the silence detection without this patch depending on the args passed to make. And we do have make 4.4 in aports now, I guess that's what lead to the bug appearing all over sudden!

So I guess old kernels without 4bf7358 will likely fail to build without V=1 now (depending on the arguments passed to make, and maybe this isn't a problem in very old kernels that have different make logic?). But also we saw that a 6.1.7 kernel failed that should have this fix already... I assume there was a regression that only got fixed later on by the refactoring commits. And it seems 6.2.0 is fine.

I'll add V=1 to the kernels that are currently failing on bpo.

By Oliver Smith on 2023-03-05T15:36:14

changed title from CI kills the build of Linux kernel to CI kills the build of Linux kernel (bug: kernel compile output silent without V=1)

By Oliver Smith on 2023-03-05T15:37:20

mentioned in merge request !3929 (closed)

By Oliver Smith on 2023-03-05T16:21:35

To summarize:

this bug appears since using a new version of make in alpine and we won't downgrade it
this should not happen anymore after upgrading your kernel to >= 6.2
if it happens beforehand, use V=1 and add this comment: # V=1: workaround for pma#1990

By Oliver Smith on 2023-03-08T08:05:25

changed the description

By Oliver Smith on 2023-03-08T07:42:35

mentioned in commit 6824414c

By Richard Acayan on 2023-03-16T08:52:51

Closing this issue, as it's a bug in the kernel sources, upgrading >= 6.2 or applying the workaround described above fixes it.

By Oliver Smith on 2023-04-02T17:08:27

closed

added status::notourbug type::bug labels

mentioned in commit 99ed2d19

By Newbyte on 2023-06-01T19:48:10

mentioned in commit 8483522c

By Newbyte on 2023-06-01T20:18:57

mentioned in commit caf3df91

By Newbyte on 2023-06-02T19:00:41

mentioned in commit 536f01be

By Newbyte on 2023-06-02T19:00:41

mentioned in commit 34b616b3

By Newbyte on 2023-06-02T19:00:41

mentioned in commit 5b2d214f

By Newbyte on 2023-06-02T19:00:41

mentioned in commit 3c7d3fb7

By Newbyte on 2023-06-02T19:00:41

mentioned in commit 7b406ace

By Newbyte on 2023-06-02T19:03:59

mentioned in commit f7e4f6e0

By Newbyte on 2023-06-02T19:03:59

mentioned in commit f296d33e

By Newbyte on 2023-06-02T19:03:59

mentioned in commit db73433a

By Newbyte on 2023-06-02T19:03:59

mentioned in commit 8a691d37

By Newbyte on 2023-06-02T19:03:59

mentioned in commit f5fc3217