[Asterisk-Dev] Bug 4301 - ztdummy accuracy problem

[Patrick]

On Tue, 2005-10-18 at 17:07 +0800, Chih-Wei Huang wrote:
> Tzafrir Cohen wrote:
> > On Mon, Oct 17, 2005 at 07:53:55AM -0400, BJ Weschke wrote:
> >> Chih,
> >>
> >>The bug was closed because the ztdummy behavior is not the specific cause
> >>for the delay problem. That patch with USE_RTC was intended to make use of
> >>the real time resource available within the kernel >= 2.6.13 instead of
> >>relying on a OHCI USB resource which was the case previously.
> > 
> > ztdummy uses the kernel's internal clock for kernel 2.6 even without
> > USE_RTC . with USE_RTC you should expect to get better accuracy, though.
> 
> I see. Thank you!
> 
> But I have further questions:
> 
> * Do I need USE_RTC for better accuracy to eliminate the delay problem?
> * Do I need kernel >= 2.6.13 to enable USE_RTC ?

Yes you need USE_RTC for better timing. And you maybe need kernel 2.6.13
to get the USE_RTC functionality because it depends on the kernel. My
FC4 box has a 2.6.12 kernel which contains the upstream USE_RTC stuff so
I patch the asterisk source to enable it on 2.6.12 too. If you use a
plain vanilla unpatched kernel than yes you will need 2.6.13.

<rant>
The delay issue in meetme has been present for a long time now and it
seems Digium is not interested in fixing it. Afaik a patch was offered
that solved the issue but was rejected by Mark because he felt the extra
thread it used would not work on embedded systems with no memory/cpu
power. Fine if it doesn't but why isn't it then possible to just do:
#ifdef normal_pc_or_server_99.999%_of_people_use
	finally_fix_meetme_delay
#endif
</rant>

Regards,
Patrick