urxvtd -f fails with EBADF

[Jukka Salmi]

Marc Lehmann --> rxvt-unicode (2007-11-27 17:07:01 +0100):
> On Tue, Nov 27, 2007 at 02:18:59PM +0100, Jukka Salmi <j+urxvt at 2007.salmi.ch> wrote:
> > Althoug the latest urxvt works fine on my NetBSD system most of the
> > time, I've occasionally seen strange things happen like urxvtd dying
> > or pututxline(3) apparently being called with an utmpx structure having
> > an empty ut_name; I can't reproduce these problems, though. Could any
> > of these be kqueue related?
> 
> Well, without further informazion its hard to say anything, for example
> which version of urxvt the "latest version" is (because that obviously
> changes over time :),

I was referring to rxvt-unicode built using the latest CVS HEAD sources
at the time of my writing.


> or at least a little bit information about how
> urxvtd is dieing (backtrace, does it even crash?),

I'm not sure. I start `urxvtd -q -o -f' from ~/.xinitrc, and urxvtc
from a script as described in urxvtc(1)'s man page, section EXIT STATUS.
I've occasionally seen

  unable to connect to the rxvt-unicode daemon: Connection refused

when running that script, thus I assume urxvtd must have died in the
meantime.


> or why you describe the
> pututxline as a problem (because that seems to be quite correct)?

Calling pututxline(3) as an unprivileged user on NetBSD results in a
setuid program (utmp_update) being called. And that program recently
logged

  Current user `jukka' does not match `' in utmpx entry

with the empty string between `' being a struct utmpx's ut_name member.
IIRC this happened while I was starting or closing an urxvtc, and
that's why I assumed urxvtc did something wrong while calling
pututxline(3).

However, I couldn't reproduce both of these problems with rxvt-unicode
built from todays HEAD sources (yet...). I'll report asap in case I
start to see the problems again.


Regards, Jukka

-- 
bashian roulette:
$ ((RANDOM%6)) || rm -rf ~