Re: Weird bugs

[Mark Burgess]

Jason, I'm not sure if you have told us what version you are using. Is
it the latest one? I am a bit sceptical about the need for such large
patches when hundred of people seem to be getting on fine.

Your message is hard to read, and is full of opinions rather than clear
facts. If there is a problem I would like to help, but I tend to ignore
long emails without clear content for lack of time. Best strategy to
prompt me into action is short and clear.

M

On Wed, 2005-10-05 at 15:02 -0400, Jason Kim wrote:
> On Thursday 29 September 2005 21:32, Jason Kim wrote:
> > I know it's bad form to reply to myself, but after much sleep and coffee I
> > feel that I can tackle this in a bit more coherent (and concise) manner...
> >
> > * There appear to be some email addressing issues with cfexecd's
> > mailing function. There was one I sent yesterday about a sscanf function
> > which is supposed to grab the domain from a 'MailTo' variable, and one I
> > found today where an 'if' statement has the same exact 'then' and 'else'
> > clauses. I've attached a patch that I believe corrects the issues.
> >
> > * The email function can (under some circumstances) use differing 'from'
> > addresses, one during the smtp 'MAIL FROM:' section and another in the
> > actual message body. Is this intentional? It doesn't seem so, in which case
> > the patch I've provided does nothing more than fix a fundamentally broken
> > method. In that case I will gladly try to neaten it up some if no one else
> > wants to.
> >
> > * The main part of my previous email was devoted to the matter of
> > extraneous runlog files caused by inconsistent handling of
> > qualified/unqualified hostnames by cfexecd and cfagent (I'm going to ignore
> > cfenvd, it's just not worth it). It appears that cfagent has the most
> > robust method of getting it right, provided that there is a 'domain'
> > variable defined. I propose patching cfexecd to instead query cfagent for
> > the correct names, as its current method of calling GetNameInfo() doesn't
> > work consistently in all cases. This could be as simple as adding 'host'
> > and 'domain' to the list of values that GetCfStuff() grabs from cfagent and
> > doing away with GetNameInfo() altogether. But if the side effects of
> > GetNameInfo() (setting various name based classes) are needed (and I didn't
> > see any sign that they were), I suppose a simple correction to VFQNAME,
> > VUQNAME, and VDOMAIN after every GetNameInfo() would work too.
> >
> > There, I hope that was better...
> > -JayKim
> >
> Replying to myself again... I'm sorry, I don't want to be annoying, but has 
> anyone had a chance (or the desire) to look into these issues? They are 
> somewhat minor, but I'm rolling out cfengine company-wide and details do 
> count.
> Would it be easier if I were to just submit a giant patch that would 'fix' 
> things the way I think they should be? I don't want to step on the Mr. 
> Burgess' toes, and I'm sure he has much more understanding of the code than I 
> do...
> -JayKim
> 
> 
> _______________________________________________
> Bug-cfengine mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/bug-cfengine