On 29/04/2024 19.49, Paul Gevers wrote:
As I hinted at in
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1069600#25, once
there's an *test* dependency relation with linux, this will be tested.
There should be a test dependency on linux-doc, but unfortunately I
cannot find any record that the uploads of src:linux to sid or
bookworm-pu triggered a test for src:dm-writeboost. Nor any trace of the
regression this should have caused. But I could be looking at the wrong
places.
All *-dkms packages in bookworm (and trixie) do have
Testsuite: autopkgtest-pkg-dkms
and with autodep8 >= 0.26 (bookworm has 0.28) this should generate for
src:dm-writeboost:
Tests: test-dm-writeboost.sh
Restrictions: needs-root, isolation-machine
Depends:
dm-writeboost-dkms,
linux-image-generic,
linux-header-generic,
stress-ng,
dmsetup,
sudo,
Test-Command: /usr/lib/dkms/dkms-autopkgtest
Restrictions: needs-root, breaks-testbed, allow-stderr, superficial,
Depends: dkms, openssl, linux-doc,
Features: test-name=dkms-autopkgtest
(The dependencies for the first test contain unreleased changes that
will try to fix the isolation-machine test, so you might see fewer deps
on the package currently in the archive.)
(The dependencies for the second test intentionally exclude
dm-writeboost-dkms because we first need to disable dkms autoinstall
before installing *-dkms (otherwise we could only get a non-verbose
package installation failure). Then we can run the module build step by
step and with improved error reporting including dumping make.log on
failure.)
I chose linux-doc because
* it is a real package built from src:linux
* it does have an unversioned name
* it is available on all architectures
* it is available on all releases since (old)+stable for more 'old'
releases than I could remember or lookup easily ;-)
Perhaps you can spot what's wrong with this setup s.t. it does not
trigger as intended.
Andreas