[Clfs-dev] Questions on SVN-20081112-PowerPC64-Multilib on ps3

[John Frankish]

At 07:36 12-12-08, you wrote:
>John Frankish wrote:
> > At 21:29 23-11-08, you wrote:
> >> According to the patch I saved (but never submitted anywhere: I
> >> figured the problem was the age of glibc), I removed the messages
> >> by changing udev/udev.h (logging_init) to remove the LOG_CONS flag
> >> from openlog.
> >
> > Thanks for the reply.
> >
> > Sorry to be dense - but if I understand correctly, you're saying that
> > I need to recompile udev after first hacking udev/udev.h as described
> > above? Note that this is with glibc-2.8
>
>Well, that's what I did.  Doesn't mean it's the right answer, of course,
>or that I actually *had* to do it.  Just that it worked.  ;-)
>
>(LOG_CONS is the flag that tells the syslog routines "write any message
>that you get to the console, if there's an error talking to syslog".
>And given the meaning, I don't know why udev doesn't normally spew all
>these messages.)
>
> > Is it possible to modify/add a start-up script to prevent this issue?
>
>I don't think so, and not that I found -- but I didn't look very hard.
>The first things I dove into were the glibc and udev sources, to find
>the source of the messages, and the reason they were being written to
>the screen instead of only the log.  (Or never-never land, which I would
>actually prefer for useless libvolume_id messages... sigh.)
>
>It *may* help to start syslogd earlier -- but when a tmpfs is mounted on
>/dev in the udev script, syslogd's /dev/log will disappear, so that idea
>may not work very well.  Also, if /var is a separate partition, there's
>no (...sane) way to start syslogd before udev in any case.
>
>
>
>_______________________________________________

I changed LOG_CONS to LOG_PID in /udev/udev.h since openlog expects 3 
parameters

..it works, thanks

..now I can see that several of the start-up scripts fail :(

John