Re: [Xen-devel] [PATCH] libxl: Make an internal function explicitly check existence of expected paths

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Tue, 2012-11-27 at 11:50 +0000, George Dunlap wrote:
> On 27/11/12 11:44, Ian Campbell wrote:
> > On Fri, 2012-11-23 at 15:56 +0000, George Dunlap wrote:
> >> +        LIBXL__LOG(ctx, LIBXL__LOG_ERROR,
> >> +                   "Internal error: Missing xenstore node %s/removable", 
> >> be_path);
> >
> > One or two of these new log lines are > 80 columns. You might find that
> > using the shorthand LOG*() macros helps with this.
> 
> Oops -- I'll fix that up.
> 
> >
> > Personally I don't think the "Internal error" tag is necessary either,
> > but maybe a new return code ERROR_INTERNAL might make sense? or even an
> > assert() if this is really a "libxl failed to maintain internal
> > consistency" ?
> 
> The idea behind the "Internal" tag is to tell users, "Unless you've been 
> manually messing around with xenstore removing nodes, this is definitely 
> not your fault."

OK. Perhaps in order to do this consistently we could have a libxl
helper which logs a message with this tag added and then aborts?
(libxl_assert?)

Might be worth leaving some time for others to weigh in before
implementing that though, since adding aborts is something which might
be considered controversial.

> I guess normally an ASSERT communicates that pretty well, but I guess 
> I'm used to ASSERTs that disappear when DEBUG is turned off. :-)  Is 
> that not the case for libxl?

I don't think assert(3) has this property?

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel