Re: KDE Bug? or an on-purpose?

[Dave Leimbach <leimySPAM2k@xxxxxxx>]

David Rhodus <drhodus@xxxxxxxxx> writes:


Thanks!  Its actually illegal to include sys/proc.h and
sys/user.h does it just fine.

This patch is very good.  I will incorporate it into whatever
other patches I have for kdebase.

Then maybe we can do an override port or something.

Dave


> Here you go.
> 
> -DR
> 
> --- ProcessList.c.orig  Wed Jul 30 00:23:10 2003
> +++ ProcessList.c       Wed Jul 30 18:59:32 2003
> @@ -28,6 +28,9 @@
>   #include <stdlib.h>
>   #include <string.h>
>   #include <sys/param.h>
> +#include <sys/proc.h>
> +#include <sys/user.h>
> +#include <sys/resourcevar.h>
>   #if __FreeBSD_version > 500015
>   #include <sys/priority.h>
>   #endif
> @@ -108,7 +111,7 @@
>           * good idea. I never thought that the stability of UNIX
> could get me
>           * into trouble! ;)
>           */
> -       unsigned int userTime;
> +       long userTime;
> 
>          /*
>           * The number of 1/100 of a second the process has spend in
> system space.
> @@ -167,6 +170,7 @@
>          struct passwd* pwent;
>          int mib[4];
>          struct kinfo_proc p;
> +       struct rusage pru;
>          size_t len;
> 
>          if ((ps = findProcessInList(pid)) == 0)
> @@ -209,7 +213,10 @@
>   #if __FreeBSD_version >= 500015
>           ps->userTime = p.ki_runtime / 10000;
>   #elif __FreeBSD_version >= 300000
> -        ps->userTime = p.kp_proc.p_runtime / 10000;
> +       if (!getrusage(p.kp_proc.p_pid, &pru)) {
> +               errx(1, "failed to get rusage info");
> +       }
> +        ps->userTime = pru.ru_utime.tv_usec / 10000; /*p_runtime /
> 10000;*/
>   #else
>          ps->userTime =
> p.kp_proc.p_rtime.tv_sec*100+p.kp_proc.p_rtime.tv_usec/100;
>   #endif
> @@ -234,7 +241,7 @@
>                          p.kp_eproc.e_vm.vm_dsize +
>                          p.kp_eproc.e_vm.vm_ssize) * getpagesize();
>          ps->vmRss    = p.kp_eproc.e_vm.vm_rssize * getpagesize();
> -       strncpy(ps->name,p.kp_proc.p_comm ? p.kp_proc.p_comm : "????",
> sizeof(ps->name));
> +       strncpy(ps->name,p.kp_thread.td_comm ? p.kp_thread.td_comm :
> "????", sizeof(ps->name));
>          strcpy(ps->status,(p.kp_proc.p_stat>=1)&&(p.kp_proc.p_stat<=5)?
> statuses[p.kp_proc.p_stat-1]:"????");
>   #endif
> 
> 
> On Dec 16, 2003, at 11:01 PM, Dave Leimbach wrote:
> 
> > The problem with ksysguardd and also apparently kcontrol is that
> > they expect certain fields to exist in
> >
> > struct kinfo_proc.
> >
> >
> > These are in /usr/include/user.h I think :)
> >
> > #ifndef _KERNEL
> >
> > /*
> >  * stuff that *used* to be included by user.h, or is now needed.  The
> >  * expectation here is that the user program wants to mess with kernel
> >  * structures.  To be sure we get kernel structures we have to define
> >  * _KERNEL_STRUCTURES.  Otherwise we might get the user version.
> >  *
> >  * This is a really aweful hack.  Fortunately nobody includes
> > sys/user.h
> >  * unless they really, really, really need kinfo_proc.
> >  */
> >
> >
> > KDE really really really really needs it :)
> >
> > The offending line looks for a p_comm field in this structure.
> >
> > if(proc_p.kp_eproc.e_ppid==ppid && p[num].kp_proc.p_comm &&
> >    0 == strcmp(p[num].kp_proc.p_comm, proc))
> >                 if(pid)
> >                     error=true;
> >                 else
> >                     pid=p[num].kp_proc.p_pid;
> >
> > I am unfortunately not sure what p_comm was in FreeBSD or what the
> > replacement is in DragonFly.
> >
> > The other field is p_runtime.  I am also not sure what the correct
> > solution is for this.
> >
> > Dave