This report is from Olaf Flebbe:
In order to reproduce:
* Use a production, _non_ DEVELOPER build of samba 3.4.x or 3.5.x
(tested it with 3.4.9 and 3.5.6)
* start smbd and nmbd.
* Use smbcontrol from 3.6.0 on the same(!) machine running nmbd/smbd
* Do "smbcontrol nmbd ping"
and you are logged out, sshd has died, X server dies
The two patches committed to master are:
These are not fixes for this issue, because it is the older binary that causes the problem, but instead fixes to ensure that at some time in the future we don't have similar issues.
However, the check for -1 may be backported, and the pull_struct_blob_all may be with care.
The only way to stop nmbd in 3.4 sending the signal will be to patch 3.4, which is now maintained for security releases only, or to have new versions (ie 4.0) follow Volker's suggestion that we move to the datagram based transport.
Issuing a new 3.4 or 3.5 release won't help many sites as these sites will probably be upgrading direct to 3.6 in any case.
In short, never mix TDB files for multiple versions of Samba at once.
Looks good, can be closed.