See http://labs.seravo.fi/~otto/mariadb-repo/mariadb-10.0-sid-amd64/lintian-0f7cb30.log and https://lintian.debian.org/tags/shlib-calls-exit.html.

The following shared libraries call exit:

• libdaemon_example.so - there seem to be no reason for exit(), use DBUG_RETURN(1) instead. (fixed)
• libmysqld.so.18 - to be analyzed
• ha_oqgraph.so - there's no exit() in oqgraph sources, most probably called by boost (grep headers for exit()) (fixed)
• ha_connect.so - either replace with abort() or keep exit(), or reassign to Oliver to handle this properly. (fixed)
• ha_innodb.so - lots of exits, either replace with abort() or keep exit(), or reassign to Jan to handle this properly.
• ha_mroonga.so - lots of exits, either replace with abort() or keep exit(), or reassign to Kentoku to handle this properly.
• ha_spider.so - either replace with abort() or keep exit(), or reassign to Kentoku to handle this properly. (fixed)
• handlersocket.so - report upstream? (fixed)

Lintian suggests:

In the case of an error, the library should instead return an appropriate error code to the calling program which can then determine how to handle the error, including performing any required clean-up.

Note: to find out if shared library has exit calls use something like nm|grep exit