Ed Swierk wrote:
> On Wed, 2009-07-22 at 20:57 +0200, Jan Kiszka wrote:
>> Nice. I noticed this, too, and shortly thought about some possible
>> fixes. Don't know why I dropped it again. Anyway, your approach looks
>> straightforward. Please rebase over git head, and I will happily test
>> and ack it!
> 
> FYI there are two other things that tend to break on laptops especially:
> 
> - If the system clock jumps backwards, functions that rely on
> gettimeofday() for computing a time interval get confused.
> gettimeofday() should be replaced with clock_gettime(CLOCK_MONOTONIC).
> Hopefully all Unixy platforms support it?

You mean that gettimeofday in updtime? Yes, that's only for timeout
calculation IIRC and should be converted. Patch welcome.

Are there more? I didn't find any on quick scan.

> 
> - Storing the address of some random host interface in our_addr is bad
> for the same reason as storing the address of the DNS server.
> Fortunately most references to our_addr are in dead code, and the
> remaining ones are questionable.

True. I shortly thought about ripping all those historic UDP protocol
emulations out, but then I didn't feel brave enough. I couldn't asses if
they are still of any use for some qemu people (namely it's CUSEEME, the
rest is disabled anyway). But maybe it's worth posting another cleanup
patch.

> 
> Anyway, here's the updated patch for the DNS issue.

Thanks, will test.

Jan