Control: forwarded -1 https://github.com/dell/dkms/issues/302
On Sun, 19 Feb 2023 17:43:50 +0100
=?UTF-8?Q?IOhannes_m_zm=c3=b6lnig_=28Debian/GNU=29?=
<umlae...@debian.org> wrote:
i strongly believe that this is not a problem with v4l2loopback-dkms,
but instead an issue with dkms itself. (i have therefore re-assigned the
bug to 'dkms').
Correct.
I've set this forwarded to a related bug ("dkms autoinstall fails on
modules not built due to BUILD_EXCLUSIVE_*")
esp. what is the use of the BUILD_EXCLUSIVE_KERNEL directive, if it
triggers a hard failure? why should i ever set the directive, if I could
just rely on the build process to fail?
IMHO, this error logic in dkms is seriously flawed. But at least we are
now at a better state than before where sometimes (always?) errors (real
errors like failing to build a module that really should be built) did
not propagate to the exit status, but only showed up as a line of text
somewhere in the middle of the (dist-)upgrade log ...
We have (had?) a piuparts report if these show up in a passing test ...
i would therefore ask the dkms maintainers to consider making
'BUILD_EXCLUSE_KERNEL' a soft-failure (that is ignored in the automatic
case).
Not trivial.
Somehow this should be treated as a tristate "OK/SKIP/FAIL".
Or two booleans
Should-be-built? "OK/SKIP"
Did_build? "OK/FAIL"
And maybe a --try-building-excluded-modules flag for those that really
want these as errors.
Andreas