Re: Random Bracket mismatch errors

[Mark Burgess]

On Mon, 2005-03-21 at 15:13 -0500, Eric Dorland wrote:
> On Mon, 2005-03-21 at 20:27 +0100, Mark Burgess wrote:
> > On Mon, 2005-03-21 at 14:10 -0500, Eric Dorland wrote:
> > > On Sun, 2005-03-20 at 09:13 +0100, Mark Burgess wrote:
> > > > Is this error reproducible? Could you get a cfagent -d0 -p dump? Is this
> > > > a recent version of cfengine?
> > > 
> > > It is not really reproducible. I see it a few times a day from random
> > > machines, but there's no real pattern to when it happens. Eric Sorenson
> > > seems to think it's a threading issue (which makes sense, given the
> > > somewhat random nature of this bug), so I'm going to try to build
> > > without threading as soon as I figure out how to do that. 
> > > 
> > 
> > Eric, please heed the following advice and answer the following
> > questions to let me help you:
> > 
> > 1) Does the error come from cfservd or cfagent or where?
> > 2) If the error comes from cfagent, you are wasting your time disabling
> > threading. Cfagent does not use threading, This will simply be a waste
> > of your time.
> > 3) Tell me what version you are using and describe the error clearly and
> > precisely in every detail. I need to see the offending lines of the
> > config file in particular,
> 
> Of course Mark, I'm sorry if I wasn't clear before and missing
> information. I'm using cfengine 2.1.13 on RHEL 3 systems, patched by
> hand to remove that SIGCHLD line (only thing I touched, I swear). The
> errors always look something like this:
> 
> cf:cfengine::/var/cfengine/inputs/cfagent.conf:2: Bracket mismatch, in
> [class=(], level = 1
> cfengine::/var/cfengine/inputs/cfagent.conf:Aborted
> 
> or
> 
> cf:cfengine::/var/cfengine/inputs/cfagent.conf:2: Bracket mismatch, in
> [class=ش(], level = 1
> cfengine::/var/cfengine/inputs/cfagent.conf:Aborted
> 
> The only thing that varies is the text after class= . It seems to always
> contain a open or close parentheses, and sometimes a character
> proceeding it.
> 
> I get this error through reports emailed to me through cfexecd, ie this
> is cfagent run through cfexecd. I've never seen this problem running
> cfagent by hand, but it's not very common that I would do that. It seems
> to happen at random times on random machines, but it happens on everyone
> of my machines (not necessarily everyday, but often enough). 
> 
> Attached is my entire cfagent.conf file, if that helps. The line 2 it's
> referring to is merely "groups:" so I don't think it's the problem. 
> 
> Thanks for any and all help.
> 
> > Without this info, I simply have no context in which to guess where you
> > error is occurring, I would like to help. but I need info.
> > 
> > best wishes
> > Mark
> > 

My guess is that this is caused by an unitialized buffer. Please try the
following patch:

http://svn.iu.hio.no/viewcvs/trunk/src/parse.c?rev=53&r1=35&r2=53