Comment 12 for bug 1028240

Barney,

On 18.09.12 11:53, Barney Desmond wrote:
> Hi there,
>
> We've been following this issue at our company and believe we have a
> firm grasp on the problem here, so I'd like to make two notes.
>
>
> 1. This bug *appears* to be related to #1002164, I'm not sure if they should be linked in any way.
>
> 2. Our description is of the problem is published here, we believe it's accurate:
> http://www.anchor.com.au/blog/2012/09/why-does-percona-cause-ssl-issues-in-postfix/
>
> Alexey, the answer to your question appears to be "yes".

Thank you very much for confirming and the detailed summary. The problem
looks clear now:

- the root cause does seem to be the YaSSL symbol conflicts with
libcrypto, which is bug #1002164.

- the same problem still exist upstream (and has been reported as
http://bugs.mysql.com/bug.php?id=65055)

- precompiled upstream packages are built against OpenSSL rather than
bundled YaSSL, so they don't exhibit this problem. IIRC only Windows
upstream packages are built with bundled YaSSL.

- our RPM packages are built with bundled YaSSL, while .debs are built
against OpenSSL.

So in addition to fixing bug #1002164 (and the upstream bug), we have to
make sure the YaSSL option in our RPM packages is intended. We might
have a packaging bug here as well.