gettimeofday seconds rollover problem?

[Joel Sherrill]

Eric Norum wrote:

> I think that we're nearing a consensus:
>
> 1) Add "memory" barrier to architecture-independent interrupt enable/ 
> disable/flash macros.
> 2) Specify "cc" in m68k interrupt enable/disable/flash macros.
> 3) Add volatile qualifiers as needed.
>
> I've committed (2).

It looks like it only got committed to the trunk.  It also needs to be 
on the 4.6 branch.

Does gcc 3.2 exhibit the movement behavior?

> Joel, will you take care of (1)?


Yes.  I have a funeral this morning but will try to get to it before 
close of business.
Was the suggestion of adding it to include/rtems/score/isr.h OK in 
conjunction
with making RTEMS_COMPILER_MEMORY_BARRIER a compiler dependent
macr?

> I guess that (3) will be dealt with whenever someone gets a chance...
>
I will double check a couple of obvious candidates but some already were 
volatile.

--joel