Hi Eric,

thanks for explaining whats' going on here, now I also understand the
need for your other "weird" dependencies.

On 2/11/25 12:32, Eric Long wrote:

`extract-linux` from kernel source code tries to do it at first (it only supports bzImage, so another custom binary runs to extract vmlinux from zImage, etc.). -cloud variant of kernel compresses with LZ4 unlike XZ
found in other variants, and `extract-linux` fails without lz4 installed
in the chroot.

That wasn't obvious from the error message I got.

To support all kernel compression methods, I included lz4, xz-utils and others in mimic-dkms' Recommends.

Seems like autopkgtest does not bring
in recommended packages,

intentionally ;-)

and xz is brought in by other packages that
Depends on it.

You don't need to care about xz-utils:
mimic-dkms -> dkms -> dpkg-dev -> xz-utils

How should I make it work? Should I include lz4 and xz-utils in Depends rather than Recommends, or is there a way to bring in extra packages in autopkgtest?

While we could add extra dependencies for the autopkgtest, I'm not sure
whether that's the right way.
I think you should promote lz4 (and any other compressor used for
official linux-image-*, but I doubt there are more) to Depends s.t. your package works out-of-the box for all official Debian kernel images.
All other decompressors that could be used by other kernel builds can
stay in Recommends.

mimic-dkms + Depends has a footprint of 43.7 MB on a really minimal
amd64 chroot, lz4 has a footprint of additional 274kb, so that is
negligible.


I'm going to try again with lz4 and on more architectures ...

Andreas

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

This is a multi-part MIME message sent by reportbug.


Followup-For: Bug #1095728

autopkgtest also works for ppc64el and arm64 in qemu chroots.
I'm updating the autopkgtest patch since we need an arch restriction
because the package is not arch:any, otherwise the tests would fail on unsupported architectures.


That oldest supported kernel seems to be 6.6, for 6.5 I get

DKMS (dkms-3.1.5) make.log for mimic/0.6.3+ds for kernel 6.5.0-5-amd64 (x86_64) Wed Feb 12 08:01:44 UTC 2025

Cleaning build area
# command: make clean KERNEL_UNAME=6.5.0-5-amd64
rm -rf build dkms.conf AKMBUILD
[ ! -d /lib/modules/6.5.0-5-amd64/build ] || make -C /lib/modules/6.5.0-5-amd64/build M=/var/lib/dkms/mimic/0.6.3+ds/build clean
make[1]: Entering directory '/usr/src/linux-headers-6.5.0-5-amd64'
make[1]: Leaving directory '/usr/src/linux-headers-6.5.0-5-amd64'

# exit code: 0
# elapsed time: 00:00:00 ----------------------------------------------------------------

Building module(s)
# command: make -j14 KERNELRELEASE=6.5.0-5-amd64 KERNEL_UNAME=6.5.0-5-amd64 CHECKSUM_HACK=kfunc
cp -rL /lib/modules/6.5.0-5-amd64/build build PATH="$PATH:/usr/lib/mimic:/var/lib/dkms/mimic/0.6.3+ds/build/../out:/var/lib/dkms/mimic/0.6.3+ds/build/../tools" vmlinux-to-btf >build/vmlinux
ERROR: no boot image found; specify BOOT_DIR and VMLINUX_SUFFIX to match $BOOT_DIR/vmlinu{x,z}$VMLINUX_SUFFIX
make: *** [Makefile:69: build] Error 1

# exit code: 2
# elapsed time: 00:00:00 ----------------------------------------------------------------

This is probably nothing that needs to be fixed.
Instead I'm documenting the fact in dkms.conf by setting BUILD_EXCLUSIVE_KERNEL_MIN="6.6". You may want to add a comment with an explanation.


Andreas

From 92a1b779da1a4926e229d2c95c981d63e93fcc54 Mon Sep 17 00:00:00 2001
From: Andreas Beckmann <anbe@debian.org>
Date: Tue, 11 Feb 2025 00:26:56 +0100
Subject: [PATCH 1/3] declare Testsuite: autopkgtest-pkg-dkms

---
debian/changelog | 6 ++++++
debian/control | 1 +
debian/tests/autopkgtest-pkg-dkms.conf | 1 +
3 files changed, 8 insertions(+)
create mode 100644 debian/tests/autopkgtest-pkg-dkms.conf

diff --git a/debian/changelog b/debian/changelog
index 111b2fd..42ea3fb 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -1,3 +1,9 @@
+mimic (0.6.3+ds-2) UNRELEASED; urgency=medium
+
+ * Declare Testsuite: autopkgtest-pkg-dkms.
+
+ -- Andreas Beckmann <anbe@debian.org> Tue, 11 Feb 2025 00:26:16 +0100
+
mimic (0.6.3+ds-1) unstable; urgency=medium

* New upstream release
diff --git a/debian/control b/debian/control
index a57da54..4998cd0 100644
--- a/debian/control
+++ b/debian/control
@@ -21,6 +21,7 @@ Homepage: https://github.com/hack3ric/mimic
Rules-Requires-Root: no
Vcs-Browser: https://salsa.debian.org/hacker/mimic
Vcs-Git: https://salsa.debian.org/hacker/mimic.git
+Testsuite: autopkgtest-pkg-dkms

Package: mimic
Archite

Hi Eric,

On 2/11/25 14:21, Andreas Beckmann wrote:

thanks for explaining whats' going on here, now I also understand the
need for your other "weird" dependencies.

The current kernel headers in sid and experimental now ship the BTF bits
and tools (e.g. linux-headers-* depends on pahole now).
Is this sufficient for your module to build without touching vmlinux to
extract the btf bits or relying on tools you ship?
If that works, I'm fine with moving lz4, pahole, ... back to Recommends.
Do you need binutils at all? It's already pulled in via dkms->dpkg-dev->binutils.


Andreas

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)