On 6/26/26 9:10 AM, Sergei Golovan wrote:
Hi Tomas,
On Fri, Jun 26, 2026 at 12:53 AM <[email protected]> wrote:
On Jun 25, 2026 17:15, Paul Gevers <[email protected]> wrote:
Hi Sergei,
On Wed, 10 Jun 2026 13:06:35 +0300 Sergei Golovan <[email protected]>
wrote:
And it appears that rabbitmq-server (4.0.5-14) currently in testing does not
work
with Erlang 29, and rabbitmq-server (4.3.0-2) currently in unstable
builds fine (with tons of warnings about newly deprecated language
features, which is annoying but harmless) but also doesn't work because
its dependency (horus) uses Erlang internal ASM format which has
been changed in Erlang 29.
For a next round of Erlang major upgrades, could you please add
versioned Breaks against known broken stuff to the erlang binaries to
ensure it doesn't migrate before the fix of reverse dependencies
migrate? rabbitmq-server was broken in testing because the fixed version
only migrated today.
Paul
It is documented upstream that rabbitmq is not yet compatible with Erlang 29.
IMO Erlang 29 shouldn't have been uploaded to Sid to begin with.
It is documented upstream that rabbitmq-server is not yet compatible
with Erlang 28 which was released a year ago. Also, upstream support
for Erlang 27 will end in May-June 2027. I don't think It's a good
idea to support an unsupported version for another 3 years. At least
now we have 5-6 months to fix bugs if there are any left.
Cheers!
Hi Sergei,
Well, what you're saying is partly truth. From upstream site:
https://www.rabbitmq.com/docs/which-erlang#erlang-28-support
"Erlang 28 is partially supported by RabbitMQ 4.3.x: upgrades of
clusters running Khepri has known issues directly related to the
breaking changes in Erlang 28 that equally affect Erlang 29."
Though upgrades have been problematic for a long time, so I don't think
this really is a big issue, it's often OK to reset a cluster after an
upgrade.
Also, after building rabbitmq-server 4.3.2, I did a few tests, and it
seem to me it was working, even with version 29 of Erlang. I may be
wrong though, as I didn't do a lot of tests (just created a queue, if I
remember well).
Anyways, it is one thing to try to have everything up-to-date. It is
another to do it without caring for reverse-dependencies.
I would have appreciate if I was made aware of the upload long enough,
and had enough time to engage with upstream to solve the issue *before*
the upload of Erlang 29 to Sid. I don't think there was such an
emergency from the Debian perspective.
Normally, this type of updates have to go trough a transition period,
and are to be dealt with the release team.
Cheers,
Thomas Goirand (zigo)
