socket_wrapper changes and ctdb

Martin Schwenke martin at
Wed Nov 22 05:06:48 UTC 2017

On Wed, 22 Nov 2017 14:58:52 +1100, Amitay Isaacs via samba-technical
<samba-technical at> wrote:

> Wow.  Now it's a different failure. :-)
> Here is the patch that fixes the test code.
> It's surprising that the tests pass without the cwrap patches
> (moving socket wrapper to third_party).  So something has
> changed in the socket_wrapper code.

Yeah, this is weird.  SIGKILL should put the process into zombie state
and, if I understand correctly, all resources should be freed (except
the process table entry).  It looks to be working like that without
socket wrapper.  However, I don't see how socket wrapper could change
that behaviour...  :-(

If I get time (though not sure I will) I'll grab the socket wrapper
tree and bisect between the versions to see what commit causes the
change in behaviour:


Extra effort in doing such a bisect is a price we pay for socket
wrapper being developed out of tree...  :-(

peace & happiness,

More information about the samba-technical mailing list