svn commit: samba r20938 - in branches/SAMBA_4_0/source/lib/events: .

[tridge at samba.org]

Volker,

 > You greatly reduced the window, but I think it is still
 > there. It's the fact that the test if we are blocked (line
 > events_signal.c:250) and the actual unblocking are two
 > different calls.

I think that one is OK, as if we are blocked (and thus go into the
if() statement) then that particular signal can't happen, so you can't
get it blocked again between the SIG_PENDING() and the unblock. 

The same logic applies when incrementing, both because sigaction()
guarantees that signals don't recurse (unless you are silly enough to
pass SA_NODEFER), and because I do the blocking before the increment.

make sense?

Cheers, Tridge