Fixes to LDB memory handling

[Matthias Dieter Wallnöfer]

It was more related to visual code analysis. I thought that if the 
message sending would fail, then we wouldn't need the message anymore 
and we could free it.

Thanks,
Matthias

Andrew Bartlett wrote:
> On Fri, 2010-12-17 at 09:08 +0100, Matthias Dieter Wallnöfer wrote:
>    
>> Andrew,
>>
>> yes, this looks to be my fault. Feel free to revert this code piece!
>>      
> Yes, I plan to.  But I still would like to know what brought you to
> making this change in the first place.  How did you determine this was a
> leak?  I would like to ensure we resolve whatever issue you were
> originally dealing with.
>
> Or was this just a the result of visual code analysis?  (i.e. reading
> the code and 'spotting' the leak).  Did you do any validation other than
> 'make test' that I should repeat after reverting the patch?
>
> Thanks,
>
> Andrew Bartlett
>
>