Hi, LeeRicky!
> Hi Sergei,
> Our storage engine writes binlog ourself to get better performance, using a
> background thread. Recently I got a bug. When shut down MySQL server, after MySQL had
> cleaned up the binlog resources, our binlog thread was still working and got into
> trouble.
> Now I want to wait for this thread to exit before MySQL cleans up binlog resources.In
> function 'cleanup()' (mysqld.cc:1295, MySQL5.1.49), I found 'ha_binlog_end()' interface
> can do this work.But it is defined as empty macro except the 'HAVE_NDB_BINLOG' macro is
> defined.
> If I implement the ''ha_binlog_end()' interface like NDB, 'binlog_func' interface of
> handlerton can be calledcorrectly and my problem is solved. But I don't want to modify
> codes in handler.cc and handler.h.
> I feel this interface hasn't completely open to storage engine. I mean storage
> engine needn't modify the codes of MySQL internal to fit requirements like mine.
> Can future version or patch solve this problem? Provide a way to let 'binlog_func'
> interface of handlertonbe called before MySQL cleans up binlog resources.
> Sorry for my poor English. I don't know if I have explained the problem clearly.
> Really looking forward to your reply. Thanks!
You can try to add -DHAVE_NDB_BINLOG to the CXXFLAGS and see if that
helps. And you can automate it by modifying CXXFLAGS from your plug.in
file.
Yes, it's a hack. For a proper solution, submit a bug report and pray
that it is fixed within our lifetime.
Regards,
Sergei
