On Thu, Sep 15, 2016 at 9:10 AM, Dodji Seketeli <dodji(a)seketeli.org> wrote:
Josh Boyer <jwboyer(a)fedoraproject.org> a écrit:
>> At the moment, the ABI changes that are reported do not trigger the
>> blocking of the build, so we need collaboration from critpath package
>> maintainers. Whenever Taskotron says "please review this ABI change",
>> the review is needed.
> Perhaps it would make sense to submit a Change for Fedora 26 to get
> this in front of FESCo and enabled as blocking. This is at least the
> 3rd time we've had an incompatible ABI change and given that we have
> the tools to prevent it, it might be time to do so.
Though, we also need to sort out how maintainers can do to say "I
reviewed the ABI change, and it's OK" -- a kind of waiving mechanism for
cases where the ABI change is harmless.
With time, we try to improve libabigail itself so that the number of
harmless changes that are reported goes toward zero. There are also
ways for package maintainers to tell libabigail to avoid showing them
classes of ABI changes that are deemed harmless; a suppression mechanism
like what the Valgrind tool has.
But ultimately, I believe maintainers need to be able to waive a
Taskotron task result that blocks an update.
Do Bodhi hackers have something in their pipeline to address this?
No. However, bodhi maintenance is changing to a new owner and now
would be a good time to start filing tickets/issues for function adds
like this. That can also be part of the Change itself, as this would
clearly be classified as a system wide change needing attention from
lots of different parts of the project.