Re: [lwip-devel] core/timers.c with NO_SYS?

[address@hidden]

Leon Woestenberg wrote:
> Hello Simon, Kieran,
>
> On Sun, Jun 12, 2011 at 2:13 PM, address@hidden<address@hidden>  wrote:
> > Aside from that, how do you propose to implement a generic timer framework
> > without knowing the time? You do need a port-dependent function to get the
> > time for that.
> Depends on whether you want to support absolute time or not. In
> previous versions, only relative time was needed (for timeouts).
That's with NO_SYS==0. And it has the downside of introducing jitter 
which depends on the number of packets coming in, up to the point where 
timeouts are not processed at all if packets come in faster than the 
stack pulls them from the queue (since the mbox code reports that it 
didn't wait at all).

There's not been such code for NO_SYS==1, however, so there are no 
previous version of timeouts for that. Now you can of course leave it up 
to the port to call sys_check_timeouts() with an argument how much time 
has passed since the last call, but that doesn't make the port easier to 
code. On the other hand, a simple function returning the current time 
should not be too hard to implement.

Anyway, you can implement it in both ways, one works is as good as the 
other. It happens to be implemented with absolute times right now since 
(to me), it's more straightforward when creating a port.

Simon