Re: Problems configuring cfengine 2.1.17 on Solaris10

[Mark Burgess]

On Mon, 2006-01-02 at 15:58 +0000, Dave Love wrote:
> Kurt Reimer <greimer@fccc.edu> writes:
> 
> > I think that all Dave Love means is that the BerkeleyDB is
> > not part of the default distribution for most sysV-like Unixes. (It IS
> > part of the standard BSD distributions, I think.)
> 
> db is in the base distribution of all systems on which I've tried to
> use Cfengine (like Solaris).  It just isn't a version against which
> cfengine will build.
> 
> >     Does the cfexecd program need to use any code from the
> > BerkeleyDB libraries in order to act as a wrapper for launching
> > cfagent or other cfengine programs? If it does and these libraries are
> > in a non-standard place, then cfexecd would not be able to tell itself
> > about that non-standard location via its "-L" switch. I think that's
> > what Dave means by "cfexecd -L isn't useful", though I'm not prepared
> > to agree at this point, not having walked through the source.
> 
> That is what I mean, but it would seem polite to check before
> concurring someone's talking a degree of nonsense.

Politeness is apparently in the eye of the beholder. I am fully aware
that I can be irritable when people fire vague and unclear aspersions
about my work. But I can also remember a time when you used to make
positive and helpful remarks about cfengine.  Let us both try to do a
better job of communicating. It seems reasonable to start with the
assumption that there is a reason for everything, however stupid you
might believe the author to be. Contexts change and things that
previously served a purpose become obsolete. -L fits into that category
for cfexecd. It was introduced because sub-shells do not always
understand the environment or the library config. Clearly the binaries
themselves have to run to begin with - but that is not guarantee that
the subshells will.

I believe the problem about building db has nothing much to do with
versions. Occasionally sleepycat change the API annoyingly, but I have
no trouble in compiling cfengine on new or old versions as long as the
libraries and header can be located. Therein lies a problem however.
Both sleepycat and package distributors seem to be at war about naming
the versions so that they might coexist. This make for autoconf
nightmare.

What would be helpful is a diagnosis of the misbehaviour, even a bug fix
or some useful information to understand the context in which a problem
is occurring. 

The gnu bug list is not my direct responsibility. I do receive some
messages from it, but I also get a lot of messages complaining about it.
I am thinking of moving away from gnu.org because there are too many
problems associated with it. I have only bad experiences of externally
hosted services.

Apologies for my failings. I hope that some simplifications will occur
with cfengine 3.

Mark