Bug#929557: Some Thoughts
On Thu, 06 Jun 2019 12:19:03 -0400 Sam Hartman <hartmans@debian.org> wrote:
>
> The linux Kernel introduced an upstream commit designed to remove an
> interface that was being misused.
No, it was removed as part of code cleanup. There was no “misuse”. It was added for KVM and existed for 7 years. KVM hasn’t needed it for a while, so patch was submitted to remove the exported function *only* leaving the GPL exported version. GKH went ahead and violated the kernel procedures and removed the symbols from LTS kernels as well. LTS is not where code cleanup should be done and symbols removed.
> That does not meet the kind of requirements for changes that we (Debian)
> make in stable releases.
> If I filed an unblock for krb5 to remove an interface at this point in
> the release process it would be outright refused.
Except this is to *not* remove an interface from LTS kernels. See above.
> We are more permissive in what changes we accept from the kernel team.
And yet you are trying to stick to the rules unlike the kernel team…
> That is, if we had the resources to review the changes adequately and do
> adequate testing, I actually suspect we would hold the kernel to the
> same standards we hold other packages to.
The symbol was exported without issue for 7 years. It’s been tested.
> I do not think this particular change would meet those standards for
> buster.
I thought LTS was supposed to be stable and not have symbols removed?
> I'm not saying the kernel team should revert the commit.
> I'm saying that the issue is far more complex than has been outlined in
> this bug so far.
>
> I think that the kernel team and the ZOL maintainers should work with
> the stable release team for buster to figure out which changes are
> permissible.
> Ultimately I'd expect that the stable release team will get to decide
> which changes they want in buster and I hope that the kernel team and
> the ZOL maintainers will work with that.
Unfortunately GKH has put his foot down with this change and insists the symbols be exported GPL only for 5.x and above. He doesn’t care much about other open source projects:
"My tolerance for ZFS is pretty non-existant. “ https://marc.info/?l=linux-kernel&m=154714516832389 except he is doing extra work to cause problems with ZFS by removing exported symbols from LTS kernels against policy.
See the pattern? No ZFS on linux user is happy about the 5.x change, but there is no valid reason to remove it from 4.x.
Also. there is no copyright violation reversing a commit, as it’s still released GPL code. GPL code is still GPL code even if it’s not in a current release. That would be interesting if code lost it’s copyright once it wasn’t being shipped.
-chris zubrzycki
- --
PGP ID: 0xA2ABC070
Fingerprint: 26B0 BA6B A409 FA83 42B3 1688 FBF9 8232 A2AB C070
========================================================
"Twice blessed is help unlooked for." --Tolkien
Reply to: