Re: careless upload of Erlang v24 without a transition tracking with the release team (was: rabbitmq-server fails to start after erlang v24 update)
Hi Sergei,
Thanks for your quick reply.
On 8/22/21 6:14 PM, Sergei Golovan wrote:
> Hi Thomas,
>
> On Sun, Aug 22, 2021 at 6:55 PM Thomas Goirand <zigo@debian.org> wrote:
>>
>> Hi Damir, Sergei, the release team,
>>
>> First of all, thanks for your bug report, Damir.
>>
>> Debian Bullseye was released on the 14th of Aug. Then Erlang v24 was
>> uploaded on the 17th. Looking at:
>>
>> https://release.debian.org/transitions/
>>
>> I cannot see any transition thingy opened for Erlang. This means that
>> Erlang was carelessly uploaded to Unstable:
>
> Uploading new major version of Erlang does not require a transition.
> No application needs to be rebuilt against it, and only a minority
> breaks (those which use removed deprecated features, and they have to
> be updated or patched anyway). I'm sorry that elixir and rabbit-mq
> break.
Here, you have 2 packages that seem to be good candidates for an Erlang
transition. If you think that's not enough packages, then probably you
could at least try to rebuild them and open bugs against the Erlang
reverse dependencies if you think that's enough? I would have been more
than happy to prepaire an upload of RabbitMQ-server before things
actually break.
>> 1/ Without informing the release team, and defining a schedule for the
>> Erlang transition
>
> I insist that a transition is not necessary.
How do you then intend that this kind of upload doesn't happen again?
Also, I don't understand your logic: what makes Erlang so special so
that it doesn't deserve the kind of care we have in other places in
Debian (ie: transitions)?
> I've uploaded Erlang 24 to experimental months ago. If you know that
> your software breaks on Erlang upgrade, you could do something
> already.
Just uploading to Experimental isn't, IMO, a thing that makes it ok to
break others unstable. For this, we have transitions... Also, an upload
to Experimental during the freeze isn't giving me any sign.
Last, I am listed as uploader for quite a bunch of packages [1] (because
I maintain OpenStack, which is big...). I know that this is a common
answer from package maintainers to say they uploaded to Experimental
first. But it doesn't work, as that's not realistic to ask to monitor
1000+ reverse dependencies for uploads in Experimental. What I do
expect, however, is having someone to file a bug against my package,
(with severity important), because that person rebuilt my package and
knows it FTBFS. And that's in fact, how transitions are supposed to be
working...
Instead, here, we received a bug report for a rabbitmq-server *user*
that discovered, after the fact, that things broke. I'm sure we can do
better than this!
Now, about Elixir, could we get it to become team-maintained, so that we
could upload new upstream releases without asking Evgeny Golyshev to do
the work? I uploaded 3 times myself, you did so 6 times, and Antonio 4
times (according to debian/changelog), compared to Evgeny 7 times.
That's IMO a very good candidate for team maintenance... Your thoughts
on this specifically?
Cheers,
Thomas Goirand (zigo)
[1]
https://qa.debian.org/developer.php?login=team%2Bopenstack%40tracker.debian.org
Reply to: