On Mon, Apr 22, 2024 at 11:07:54AM -0000, Matthieu Baerts wrote:
> > I don't think we can justify dropping FAN support in an SRU, so let's
> add that port back in.
>
> Even if it will not be used in Ubuntu 24.04,
I'm happy to drop it if the stable release team would accept it.
However, I don't see how we can argue that dropping FAN support is
*not* a regression, even if the kernel we ship does not support it.
> and dropping these patches would stop the diversion with Debian?
That's not a goal of our SRU process. We can always realign with
Debian during the next devel cycle.
> > @matttbe - is there any documentation about upstream compatibility guarantees and regression testing done that you're aware of, or any tests that we could run beyond those in our autopkgtests?
>
> @dannf: IPRoute2 is following the kernel development, and it is not
> supposed to break the compatibility with older kernel versions (also
> because the kernel's user API is not supposed to break).
>
> There is a test suite included in the source code, but it looks like it
> is not tested when building the .deb package, check the 'debian/rules'
> file:
>
> override_dh_auto_test:
> # upstream test suite needs root and leaves machine unclean, skip it
The package does seem to include this in its autopkgtests
(debian/tests/testsuite.sh), so I think we have that coverage.
>
> I never used it, (and it looks like it is not often updated), but IPRoute2 (ip, tc, ss, etc.) is also heavily used in the kernel selftests:
>
> $ git grep -cw -e ip -e tc -e ss -- tools/testing/selftests/ | wc -l
> 635
>
> Maybe these tests are already being validated when building a new kernel for Ubuntu? That's another way to validate IPRoute2.
Thanks for the pointer!
-dann
> Other than that, Debian is using IPRoute2 versions > 6.1 for a long time
> now.
>
On Mon, Apr 22, 2024 at 11:07:54AM -0000, Matthieu Baerts wrote:
> > I don't think we can justify dropping FAN support in an SRU, so let's
> add that port back in.
>
> Even if it will not be used in Ubuntu 24.04,
I'm happy to drop it if the stable release team would accept it.
Here's the policy: /wiki.ubuntu. com/StableRelea seUpdates
https:/
However, I don't see how we can argue that dropping FAN support is
*not* a regression, even if the kernel we ship does not support it.
> and dropping these patches would stop the diversion with Debian?
That's not a goal of our SRU process. We can always realign with
Debian during the next devel cycle.
> > @matttbe - is there any documentation about upstream compatibility guarantees and regression testing done that you're aware of, or any tests that we could run beyond those in our autopkgtests? dh_auto_ test:
>
> @dannf: IPRoute2 is following the kernel development, and it is not
> supposed to break the compatibility with older kernel versions (also
> because the kernel's user API is not supposed to break).
>
> There is a test suite included in the source code, but it looks like it
> is not tested when building the .deb package, check the 'debian/rules'
> file:
>
> override_
> # upstream test suite needs root and leaves machine unclean, skip it
The package does seem to include this in its autopkgtests tests/testsuite .sh), so I think we have that coverage.
(debian/
> selftests/ | wc -l
> I never used it, (and it looks like it is not often updated), but IPRoute2 (ip, tc, ss, etc.) is also heavily used in the kernel selftests:
>
> $ git grep -cw -e ip -e tc -e ss -- tools/testing/
> 635
>
> Maybe these tests are already being validated when building a new kernel for Ubuntu? That's another way to validate IPRoute2.
Thanks for the pointer!
-dann
> Other than that, Debian is using IPRoute2 versions > 6.1 for a long time
> now.
>