Timer "wraps"?

Scott M

It looks like the timer class implements a 32 bit timer. This means the timer wraps every 10.5 hours or so. (I'm rounding off to keep the numbers simple, here).

This should raise all sorts of problems in my application. If it's been running for, say, 10.4999 hours, and I then set a timer for 10 minutes, I'm going to guess that things are going to fail, badly. The math will wrap and the very next timeout check I do is likely to return true. Is this correct?

I can think of ways to write the Timer class differently, that would avoid this problem (well, mostly: there is no simple fix that doesn't make at least some assumptions.) Is this a known problem, and has anyone already come up with a fix?

Peter Verkaik

No, the overflow does not pose a problem, because the

timeout() method calculates the differences with respect

to the current timer value.

You only have a problem if you want a timeout longer than

the 10.5 hours. Then you must keep the time in a

private variable. For example, if you want a timeout for

24 hours, you should test for a timeout of 1 hour, and

then increment your private hour counter. Then, when

that reaches 24, clear it and execute your code you

want to run once every 24 hours.

The problem with those longer timeout values is the

required 32bit integer math. Normally, with longer

intervals you don't need the 8.68 microsecond resolution

of the 32bit timer. For that purpose I have written a 16bit

timer class that allows selectable time units up to 0.6 seconds

per tick (65536 * 0.6 sec = your 10.5 hour).

Check out my ScaledTimer16 class here:

http://groups.yahoo.com/group/JavelinCode/files/Javelin%20Stamp%20IDE/lib/stamp/core/

regards peter