Do not free the header while unmapping the xxxx_UNPREPARE messages

[Dmitry Timoshkov]

"Eric Pouech" <eric.pouech at wanadoo.fr> wrote:

> I'd rather see the fix being made completely in message16.c by adding 
> the return value to the unmapping functions

This approach doesn't look clean to me.

> IMO, it better encapsulates all the 32<=>16 message mappings, and it's 
> bad to spread this across several compilation unit

Freeing the header while unmapping the 16/32-bit message is an alien task
for mapping functions IMO.

-- 
Dmitry.