non-responding processes after truss(1)ing

[Jeremy Chadwick]

On Tue, Sep 27, 2011 at 09:28:56AM -0400, Mark Saad wrote:
> On Tue, Sep 27, 2011 at 6:19 AM, Nikos Vassiliadis <nvass at gmx.com> wrote:
> > On 9/27/2011 1:10 PM, Jeremy Chadwick wrote:
> >>
> >> kill -9 your truss processes; the underlying processes which you are
> >> truss'ing will probably resume.
> >>
> >> My experience for years has been that truss on FreeBSD is extremely
> >> buggy and cannot be relied upon (case in point). ??Such is still the case
> >> on RELENG_8 as of today.
> >>
> >> Use ktrace(1) instead. ??You'll find it to work pretty much in every
> >> situation.
> >>
> 
> What about using dtruss in place of truss is the dtrace implementation
> of truss any better then the old libkvm ?

This pulls in a whole can of worms.  Getting DTrace to work on FreeBSD
is a little tricky, because certain commands/arguments must be provided
manually during world/kernel "make" time and not via make.conf/src.conf.
AFAIK this is still the case in RELENG_8, while "kludges and hacks" have
been put in place on 9.x to work around this.  I can provide some
references to my claims if need be.

There's also some segregation between DTrace-capable userland and
DTrace-capable kernel, but the delineation between the two -- and how to
accomplish one without the other -- is something I've never found any
conclusive write-up on or otherwise.  I think such an explanation would
benefit many userland application authors/developers.

-- 
| Jeremy Chadwick                                jdc at parodius.com |
| Parodius Networking                       http://www.parodius.com/ |
| UNIX Systems Administrator                   Mountain View, CA, US |
| Making life hard for others since 1977.               PGP 4BD6C0CB |