[Pkg-erlang-devel] Bug#591431: erlang 14.a breaks ejabberd

[Sergei Golovan]

reassign 591431 ejabberd
thanks

On Tue, Aug 3, 2010 at 1:46 PM, Marc Dequènes (Duck) <duck at duckcorp.org> wrote:
> Coin,
>
> Quoting Sergei Golovan <sgolovan at nes.ru>:
>
>> Please, check if ejabberd 2.1.4-1 from sid works with erlang
>> 1:14.a-dfsg-2.
>
> I forgot to say i tested 2.1.4-1, and i'm now using it with erlang
> 1:13.b.4-dfsg-6 successfully.

OK. Bad to hear that.

>
>> erlang-abi-13.a means that binary interface to C programs is compatible
>> with
>> Erlang R13A, which is true. So, it isn'r the bug source.
>
> That was to check if the provides field was correctly reflecting reality.

I see.

>
> I suppose the code in question is related to X509 certificates used for SSL,
> but i see no other error message in the log except for an unrelated minor
> pubsub problem and all SSL connections work well (c2s, s2s, ejabberd
> web-interface with SSL).
>
> It may also be related to LDAP, but authentication works perfectly.

I've conculted with one of ejabberd authors who says that these warnings are
harmless, so the problem seems to be somewhere deeper.

>
> I was using ejabberd 2.1.3-2 with 1:13.b.4-dfsg-5 on another node and saw no
> such problems. I upgraded to 2.1.4-1 with the same erlang version and it's
> ok. I may try to upgrade erlang on this node at night to try to reproduce
> the problem.

ejabberd was known to break on erlang 14, and there was at least one patch
applied to make it build on erlang 14, so may be it's another incompatibility.

>
> If you think this is not a widespread problem, feel free to downgrade
> severity. I just don't know what would be the next step to debug, so please
> help me if you can.

I've contacted ejabberd maintainer who agreed that it's likely to be an ejabberd
fault, so now I'm reassigning this bug to ejabberd. If it'll turn out
that the bug in
Erlang feel free to reassign it back.

Cheers!
-- 
Sergei Golovan