Re: bzr commit into mysql-5.1-bugteam branch (Dao-Gang.Qu:3192) Bug#42851

[Davi Arnaut]

On 11/26/09 7:32 PM, Andrei Elkin wrote:
> Davi,
>
>     >  You have a point. But, Davi, we were "reflecting" today at a phone call
> chatting with
>     >  Serge today.
>     >
>
>     Serge? You mean Sergei?
>
> right (sorry, I thought I cc-ed him).
>
>
>     >      The problem is that a
>     >      user might trigger messages to be written to the error log, which is a
>     >      concern since disk space for the error log is limited. The concern is
>     >      about messages written to the error log.. any kind (be it a warning, a
>     >      error or a note).
>     >
>     >  It's a problem, but still separate to one of the bug#42851.
>
>     No. The original problem ("spurious" foo messages) of Bug#42851 has
>     already been solved, a patch was pushed and everybody was happy. But
>     Domas raised a point that the DoS (flooding of the error log) was also
>     a concern and needs to be solved. Hence, the bug was kept open and
>     this is what needs to be addressed now.
>
>     >      This has nothing to do with push_warning nor it is
>     >      exclusive to replication code -- the case reported on the bug is just
>     >      one of the various that a user can trigger.
>     >
>     >  In our consensus view the one is about unfortunate style of sending a
> particular replication
>     >  warning to the user through the error log instead of the session warnings
> stack.
>
>     Yes, but this is solved already.
>
> I have to admit, that I really forgot that a patch was pushed. Sorry.
> However, my previous mail was saying happiness is not any longer true.
> I mean lines of the original "Description:
>     Since upgrading a development system to 5.1.31, I have noticed the following
> kinds of
>     warnings in the error log:
>     090213 16:58:54 [Warning] Statement is not safe to log in statement format. ..."
> refer to a warning that probably should not have been in the error in the first
> place.

Sergei G. (in a e-mail to Lars with subject "Re: BUG#42851: How to treat 
warnings?) explains in detail why this also goes to log file (but, 
attention, this only happens if log-warnings is enabled -- consequence 
of the pushed patch).

>
>     >    From point of view of the error-log flooding, this view would be looking
> as a workaround.
>     >  Anyway, with or without fixing the flooding issue, our UNSAFE warnings
> should not go
>     >  into the error-log.
>
>     And it has already been solved. The issue at hand is the
>     flooding. This is specifically why the bug has been left open.
>
> That's understood. What's not is that it would be reasonable to open a new bug
> report.

Perhaps it would be better to open a new bug report.

> I am puzzled how to proceed.
>
> #1, I would like to talk to repl team once again about converting the pushed patch
> for bug#42851
> into one that would deploy push_warning().
> #2, I would be happy to contribute to the flooding issue whatever means I am capable,
> but
> let's wait for the #1.
>
> Ok?

The push_warning is already deployed. What the code does is both, it 
pushes a warning to the statement stack (push_warning) and also sends a 
warning to the log file (if log_warnings is enabled).

I think we should focus on #2.

Regards,

--
DAvi Arnaut