Re: Adding makemandb as build tool for man.db
David Brownlee <[email protected]> Tue, 16 Oct 2018 15:55:53 +0100
| Newsgroups | gmane.os.netbsd.devel.installation,gmane.os.netbsd.ports.mac68k |
|---|---|
| Message-ID | <CAGN_6pZ3eju_u8gJmyha7V9oHo8aycObWb86z2X1HMOWJodw-g@mail.gmail.com> |
On Wed, 10 Oct 2018 at 20:05, Robert Swindells <[email protected]> wrote: > > [email protected] wrote: > >On 10/9/18 3:02 AM, Martin Husemann wrote: > >> On Tue, Oct 09, 2018 at 09:46:48AM +0100, David Brownlee wrote: > >>> On lower powered (and particularly lower memory boxes), the makemandb > >>> on first boot after install can run for some time and significantly > >>> affect the performance of the system (in extreme cases for several > >>> hours). > >> It is not so much memory, but slow disk access that makes it painfull. > >> Most sparc machines do pretty well, mac68k is a disaster. > >> > >> Even if you distribute the database, it will take some serious time on > >> each boot to check it is up to date. > >> > >> Maybe we should add an rc.conf option to delay it after boot (i.e. sleep > >> three hours and only then start the man page check), defaulting to 0 (i.e. > >> no delay) on most architectures but make sysinst configure it to a few > >> hours on slow architectures. > > > >Maybe add a rc.conf option to disable man page processing > >altogether? > > We already have one: > > makemandb=NO > > You also need a line in /etc/daily.conf though to keep it turned off: > > run_makemandb=NO As a useful initial step how about sysinst setting makemandb=NO for low memory/cpu machines (possibly RAM < x or uname-p in y), we can then expand that later. I would be inclined to leave the run_makemandb alone in cron at this stage as its the overnight processing, but happy to be persuaded otherwise :) David