Bug#490859: bug #490859: libaprutil1 Depends libmysqlclient15off >= 5.0.27-1 is wrong

[Stefan Fritsch <sf@sfritsch.de>]

On Tuesday 15 July 2008, you wrote:
> From: "Stefan Fritsch" <sf@sfritsch.de>
> Sent: Tuesday, July 15, 2008 10:11 PM
>
> > What do you mean with defunct? Does top show the process as
> > "defunct" or is it just doing nothing? If the latter, could you
> > please provide a backtrace, as described in
> > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=480210#10
>
> # ab -n 10000 -c 1000 http://localhost/ &>/dev/null && ps aux|grep
> defunct www-data  2105  0.2  0.0      0     0 ?        Z    22:21  
> 0:00 [apache2] <defunct>
>
> I needed to run that command twice now to get one defunct process
> Tried gdb -p on it and noticed now a weird thing
> Did again ps aux|grep defunct and the pid was different.
> If I just press key up and then do the ps aux|grep defunct again
> and again I almost everytime got a new pid
> I thought defunct processes can only be removed by a complete
> system reboot As least I remember that I had that once months or
> even years ago.

I expect this happens only during the ab run and the defunct processes 
disappear when ab is finished. This is not a bug, then. Processes 
that have terminated but for which the parent process has not yet 
retrieved the exit status are shown as defunct. Under high load, this 
can take some seconds.

>
> > This should be the case.
>
> Oh noticed I forgot to mention that this would mean the number
> hasn't bumped since libaprutil1 was last uploaded.

I think there is a misunderstanding here. The dependency will be set 
during the compilation of apr-util by the shlibs file in the 
libmysqlclient15off package that was used to compile apr-util. The 
current -dev package still lists 5.0.27-1 in its shlibs file, so that 
should work. If it doesn't, that would be a bug in 
libmysqlclient15off. But as stated above, I don't think there is a 
problem.

Cheers,
Stefan