the sorry saga of the talloc soname 'fix'

Michael Brown mbrown at
Wed Jul 8 16:14:15 GMT 2009

On Wednesday 08 July 2009 09:15:52 Sam Liddicott wrote:
> <snip>
> And, (here I really do start to repeat myself) we get all of this merely
> by:
>    1. preventing the internal promotion of reference to parent
>    2. using a "no_owner" place holder as parent when talloc_free is
>       called while references are still held. This no_owner does not
>       have the power to prevent destruction when the references all go.
>       This no_owner can be stolen from.
> The patches for this, including patches to the test suite can be found in:

Score: +5, Actually solves problem.

I've been following this thread with interest, discovering a world of shared 
library pain that I never knew existed, all of which seems tangential to the 
real solution of making the API unambiguous.  I wouldn't normally jump into a 
discussion on a project I don't actively contribute to, but it would be very 
sad to see such an elegant fix drown in the noise.


More information about the samba-technical mailing list