fetchmail service do not stop cleanly;it ignores TERM signal

[Christoph Brinkhaus]

On Tue, Mar 29, 2016 at 11:34:29PM +0200, José García Juanino wrote:
> Hi,

Hello José,
> 
> I have noticed that fetchmail service hangs in stop phase. I have the
> following in rc.conf:
> 
> fetchmail_enable="YES"
> fetchmail_users="juanino"
> fetchmail_juanino_polling_interval=60
> 
> 
> When I have to restart the service, it hangs forever:
> 
> # service fetchmail restart
> ===> fetchmail user: juanino
> Stopping fetchmail.
> Waiting for PIDS: 9284
> 
> The only way to stop the service is to send the INT (or KILL) signal:
> 
> # pkill -INT fetchmail
> 
> It seems to be that is ignoring TERM (or HUP) signal.

I have faced the same situation.
> 
> Ugly workaround:
> 
> echo "sig_stop=\"INT\" " > /etc/rc.conf.d/fetchmail
> 
> to force the INT signal instead of TERM.
> 
> The same issue applies to some other services as polipo. It is not
> clear for me if this issue is freebsd specific. Any idea?

Some days ago I have seen a related report about 10.3 on the stable
mailing list. It has been reported that some services to not stop
properly and the enabling of SAVESIGVEC has been detected as a likely
trigger. If I remember correctly the proposal was to disable SAVESIGVEC
until the real root cause is fixed.
may be in the current kernel sources it is already fixed or SAVESIGVEC
is re-disabled.

I have digged a little bit and found SAVESIGVEC defined in
/usr/src/bin/csh/config_p.h. I have simply commented that out 
and made a new world and a new kernel. This is my current work
arround. As said above I am not sure about the current situation.

I hope this information helps.

Kind regards,
Christoph