Tom Lane wrote:
> Magnus Hagander <magnus@hagander.net> writes:
>> Not likely, but I'd test it anyway. If the issue is related to AV, it's
>> certainly fine - you won't be running AV on your Solaris. But more
>> importantly, Unix has actual support for signals and not just the fake
>> stuff we have on Win32, so it's likely that the postmaster will be
>> capable of killing the child processes.
>
> I'm not sure what failure mode you're imagining, but the postmaster has
> already verified that all the children that are supposed to be connected
> to shared memory are dead before it attempts to recreate shared memory.
> So the above sounds completely bogus.
>
> I'm still suspicious of the syslogger holding onto an inherited handle
> to the shared-memory file, though that theory would seem to mean that
> crash recovery would never work at all on Windows if the syslogger
> were enabled. But maybe there is some additional gating factor needed
> to cause the problem to manifest.
Well, the syslogger is enabled by default on *all* binary installs on
windows, so I think we would've seen more if it never works.
I'll see if I can repro a case like it to see if the syslogger prevents
the shared mem from going away when I get back to a dev box. Should be
enough to just stick a sleep preventing it from stopping, right?
//Magnus