Just a quick thought - would using "nice" to run htdig at
a lower priority be a possible workaround to slow down
the indexing?
Gilles Detillieux wrote:
>
> According to Geoff Hutchison:
> > At 12:08 PM +0000 3/23/01, Malcolm Austen wrote:
> > >of >10,000 pages to be indexed from one server I just can't manage with
> > >only integer seconds control.
> > >
> > >Could an alternative be considered please?
> >
> > Well, the problem is one of system calls. Right now, we're using
> > sleep() which only takes integers. I would guess the biggest
> > stumbling block to using something else is whether there's a very
> > portable alternative. I can see calls for setitimer() on my Linux
> > box, but I don't know if this is actually part of POSIX or not.
>
> My vote would be to stick with server_wait_time, and make it a Double
> rather than integer value, where the handling of fractional values would
> be system-dependent.
>
> Linux has nanosleep(), which conforms to POSIX.1b (formerly POSIX.4),
> and uses a structure that defines both seconds and nanoseconds to sleep.
> It also has usleep(), which conforms to BSD 4.3, and accepts a single
> unsigned long int specifying microseconds to sleep (this would work as
> long as server_wait_time < 4280 or thereabouts). Both have a granularity
> of 1/100 s on Linux x86, which should be adequate for this purpose.
>
> To be complete, we should have configure test for both, and probably use
> nanosleep() by preference, if available. To be really thorough, I suppose
> we could test for setitimer() on systems that have neither nanosleep()
> nor usleep(), which might help on some non-POSIX.1b conforming SysVr4
> systems, but this would require signal handling and would get a bit messy.
> I suspect Linux implements both nanosleep() and usleep() library functions
> using setitimer() and sigaction() systems calls. It's probably a safe
> bet that any system that has setitimer() will also have nanosleep()
> or usleep(). (Someone correct me if I'm wrong!) Indeed, if recent
> BSD systems have nanosleep(), we may need to implement fractional
> server_wait_time values using only that, and not bother with usleep().
>
> --
> Gilles R. Detillieux E-mail: <[EMAIL PROTECTED]>
> Spinal Cord Research Centre WWW: http://www.scrc.umanitoba.ca/~grdetil
> Dept. Physiology, U. of Manitoba Phone: (204)789-3766
> Winnipeg, MB R3E 3J7 (Canada) Fax: (204)789-3930
>
> _______________________________________________
> htdig-general mailing list <[EMAIL PROTECTED]>
> To unsubscribe, send a message to <[EMAIL PROTECTED]> with
>a subject of unsubscribe
> FAQ: http://htdig.sourceforge.net/FAQ.html
--
Tim Peterman - Web Master,
IT&P Unix Support Group Technical Lead
Lockheed Martin EIS/NE&SS, Moorestown, NJ
_______________________________________________
htdig-general mailing list <[EMAIL PROTECTED]>
To unsubscribe, send a message to <[EMAIL PROTECTED]> with a
subject of unsubscribe
FAQ: http://htdig.sourceforge.net/FAQ.html
