utmp : just when you thought it was safe...

[David Lee]

F.A.O. Jeremy and those exploring the murky corners of utmp ...

It always happens, doesn't it?  The project is nicely wrapped up, all
seems settled, and then someone says "I think there's a problem".

Executive summary:  there is a minor bug in a post-2.0.7 patch to utmp
handling.

Detail:  2.0.7 introduced utmp support on an experimental basis.  Since
then I have been developing a patch to improve its portability and
functionality.  A few weeks ago, everything had stabilised.  Great!  Then
last night I received an email from someone whose smbd was crashing, and
the report had "utmp" close to the point of failure.  My heart sank... 

I had a quick look at the source code of 2.0.7+patch, and found that my
patch introduced a bug into connection.c/"utmp_claim()": it is now using
variable "host" before setting it.

I'm not aware that this is causing any problems, but it needs fixing. 
And in particular, I'm not sure whether this causes my correspondent's
bug, or whether it is just happy coincidence.

Unfortunately I'll be unable to do any debugging over the next few days,
although I have meanwhile indicated to my correspondent the problem area,
should he wish to chase my bug and its possible link to his problem. 

Meanwhile:

Jeremy:
(a) Can you indicate whether you rolled my patch from a few weeks ago into
    CVS?
(b) In what form should any future fix be submitted: by adjusting my
    earlier patch (against "neat" 2.0.7) or as standalone against any
    already-updated version?
(c) My apologies for this extra hassle!

-- 

:  David Lee                                I.T. Service          :
:  Systems Programmer                       Computer Centre       :
:                                           University of Durham  :
:  http://www.dur.ac.uk/~dcl0tdl            South Road            :
:                                           Durham                :
:  Phone: +44 191 374 2882                  U.K.                  :