Alfranio, hello.
> Hi Andrei and Jasonh,
>
> I've implemented what follows as we have discussed yesterday:
>
> 1 - The decision on what type of cache to use is delegated to
> the log.cc file, specifically to MYSQL_BIN_LOG::write.
>
> 2 - In order to temporary identify a DDL, I've used
> CF_AUTO_COMMIT_TRANS as suggested by Jasonh.
>
I still think it would be better if DDL callers marked THD instance
specifically. However, my objection is not critical.
> 3 - I am not calling thd->binlog_start_trans_and_stmt() when
> writing directly to the binary log, e.g. DDL.
>
> 4 - To further influencing what type of cache to use, I've introduced
> a flag (NO_CACHE) which is stored along with the flag is_transactional
> in a bit map.
>
> 5 - I also addressed some other minor details requested by Andrei.
>
More important is to provide comments as in the code - e.g the
heading synopsis lines for new functions - as well as per file
comments in the cset.
> I am running mtr to check if everything is ok.
> I will commit a patch at the end of the day and ping you both.
>
>
I am ready to look at it today.
good luck,
Andrei
> Cheers.
