On Fri, Mar 04, 2005 at 11:12:59AM -0600, Luke Kanies wrote:
>
> The only real changes I made, other than to libmetrics.{c,h}, was to
> move the metric_init() function to the end of each metrics.c file (so
> that it is aware of each of the metrics functions) and call new_metric()
> for each metric. I could have also just provided a prototype for each
> metric and left metric_init where it was, but this seemed easier in the
> long run.
Please use prototypes. That way there will be a list of metrics at the
top of the file and you won't have to scroll through the thousands of
lines of code to find them all. If you do that and keep metric_init()
at the top, the compiler will helpfully complain when you forget to
update the list.
> I also set up test-metric so that it uses the dynamically-built list of
> metrics, so it should be able to test each platform's metrics pretty
> easily.
Great!
> You can get access to the code at http://madstop.com/svn/libmetrics
>
> Please take a look and provide any comments you might have. I'd like to
> get this code put into the main tree.
>
> I hope to start incorporating the metrics that cfengine collects, also,
> and then convert it to just use libmetrics for its collection. This
> would be much easier, though, if libmetrics were distributed outside of
> ganglia. Would it be acceptable to make this available outside of the
> ganglia package? Lastly, the license is currently GPL, but it makes
> sense (I think) for library licenses to be LGPL -- in externalizing the
> package, would it be acceptable to make that change in license?
Hmm, this is a bit odd. The ganglia license is BSD, is libmetrics
really supposed to be GPL? If its, you might as well GPL the whole
thing.
-- Brooks
