Re: Tcl test hangs 5.0_BETA

["Andrew Sporner" <asporner%googlemail.com@localhost>]

I had this problem in 4.0 on my MVIII.  I had thought it was the SCSI
controller (which was a little odd to start with) but this makes sense too.

My network sessions hung.  But this is a general interrupt problem, that
would make sense.


Andy


On Fri, Jan 16, 2009 at 6:01 PM, Johnny Billquist <bqt%softjar.se@localhost> 
wrote:
> Hans Rosenfeld wrote:
>>
>> On Fri, Jan 16, 2009 at 08:12:13AM +0100, Anders Magnusson wrote:
>>>
>>> Hm, can it be that the console output driver just hangs sometimes?
>>> That has happened before, and jumping into the debugger and back solves
>>> the problem.
>>
>> I just had a "hang" in a native build. Just entering and leaving the
>> debugger didn't help, but single-stepping around a bit (into the console
>> interrupt handler) got the system out of the "hang".
>>
>> Johnny, could this be the problem you saw with long-running native
>> builds?
>
> So it would seem. Stepping on a little in ddb, and then continuing got it
> running again.
>
> My build was aborted, but the machine is alive again.
> (The build stopped with a signal 5, well lint did.)
>
> Interesting. So it appears we have some problem with interrupts being locked
> out, and eventually not even being restored at all.
> I seem to remember something about a change to the splxxx() calls a while
> back. That could be the problem here...
>
>        Johnny
>
>