From: Ann W. Harrison
Date: September 25 2008 7:03pm
Subject: Re: falcon memory leak
List-Archive: http://lists.mysql.com/falcon/2
Message-Id: <48DBE09A.9010206@mysql.com>
MIME-Version: 1.0
Content-Type: text/plain; format=flowed; charset=ISO-8859-1
Content-Transfer-Encoding: 7BIT

Hi Kelly,

> 
> 
> Ann.  You should look at the log file too... I'm seeing invalid access 
> of memory in the serial log file handling.  It may not be an issue but I 
> know you are chasing a recovery issue and this could be the initial 
> trouble that leads to recovery issue.
> 

I think not.  The first three are in findLastBlock which reads the
serial log file until it gets a block that doesn't look healthy -
either the number or timestamp is wrong.  When the last window was
not completely filled, the last good block in the window is followed
by uninitialized memory.  We could avoid this error by zeroing the
window before reading, but that doesn't seem all that efficient.
We do zero the window if HAVE_purify is defined.  Maybe we need to
zero the window more often.

The other errors are in write, where we round up from the last
block to be on a sector boundary, so we write parts of the buffer
that we haven't initialized.  They're the other side of the first
problem.

I'll look for a safer way to check for the end of the written
window.


Cheers,

Ann