gettimeofday seconds rollover problem?

[norume at aps.anl.gov (Eric Norum)]

On Feb 24, 2006, at 8:43 AM, Joel Sherrill wrote:
> I agree with everything above.  I do not want to add unnecessary  
> register dependencies.
> Plus this technique should be able to be added fairly mechanically  
> to the other ports.
>
> The next question is what other SuperCore variables need to be  
> volatile.  _Thread_Executing,
> _Thread_Heir, and _Context_Switch_necessary seem very high on the  
> list.

Yes, I think that these should also be volatile.  Their values can be  
changed by another thread so the compiler must be informed of this.

>
> Eric.. will the EPICS team submit a fix for the m68k, i386, and  
> powerpc ports?  I can probably
> do the others myself based upon those.

I'll commit the changes for the m68k which I proposed in the bug  
tracker yesterday.
Till, will you do the powerpc?
I've not done much x86 assembly programming.   Anyone else like to  
take a crack at this one?

-- 
Eric Norum <norume at aps.anl.gov>
Advanced Photon Source
Argonne National Laboratory
(630) 252-4793